Residential Iatrogenic Effects i Running head: RESIDENTIAL DEVIANT BEHAVIOR IATROGENIC EFFECTS

Youth in Residential Treatment: Deviant Behavior and Iatrogenic Effects in Peer Groups

A thesis presented by Caitlin M. Delaney to the Connecticut College Department of Psychology in partial fulfillment of the requirements for the degree of Bachelor of Arts

Connecticut College New London, CT May 1, 2009

Residential Iatrogenic Effects ii Acknowledgements First and foremost I would like to thank Dr. Audrey Zakriski for her continuous support and constructive feedback. Professor Zakriski, I really appreciate your enthusiasm all year about my work, and how much you have encouraged me when feeling overwhelmed. Even while away on sabbatical, you have found the time to read and revise troubling sections of my thesis five or more times. To my readers, Dr. Jefferson Singer and Dr. Jennifer Fredricks, thank you for your feedback and encouragement. To Wediko Children’s Services, for providing me with the opportunity to become part of such a great program that truly makes a difference in children’s lives. Thank you to the Wediko Summer Program staff, particularly the Otters, for understanding my love for research and supporting me through summer and beyond. To all the children of Wediko who have not only inspired my work, but who have also provided countless memories that have made the summer of 2008 one of my best. To Dr. Stuart Vyse for teaching me as much as possible about SPSS, and for making me truly appreciate statistics from t-tests to MANOVA’s. Thank you for sharing your knowledge and your support since freshman year. Also to my friends at Connecticut College, from Andover, and from Ireland, who have all encouraged me to keep writing all year and have been supportive through every step of the process. I would also like to thank my family. To my Mom and Dad for their unconditional love and support and for giving me the opportunity to attend Connecticut College. Also to my brothers Jimmy and Matt for reminding me to take advantage of being in college.

Residential Iatrogenic Effects iii Abstract Aggregating antisocial youth in intervention programs has been common practice for years. Group therapy is more efficient and cost effective than individual therapy, and it has also been seen as desirable because of the opportunity it provides for peer feedback and support. Despite these benefits, researchers have recently begun to observe and examine the negative effects of aggregating deviant youth for group treatment. Specifically, this research has shown that group treatment has the potential for unintended, negative effects (iatrogenic effects) that occur as peers reinforce each other’s antisocial and deviant behaviors. Much of the research on iatrogenic effects focuses on community-based treatment, school settings, and juvenile detention centers. This study examined iatrogenic effects in residential treatment, paying particular attention to the child’s externalizing and internalizing behaviors, as well as child-group goodness-of-fit. Data were collected at several points over the course of a 45-day treatment program. Assessments rated the child’s internalizing and externalizing behaviors, as well as how they interacted with peers and whether they reinforced different types of behavior. Results suggested that while mixed and externalizing children showed problematic behavior patterns including elevated antisocial and decreased prosocial behavior compared to non-externalizing youth, externalizing youth more actively encouraged antisocial norms within the peer group. When group composition was examined, it was shown that aggregating externalizing youth led to increases in antisocial influence over time, regardless of child type. Regarding child-group fit, however, non-externalizing youth showed the clearest evidence of iatrogenic effects due to deviant peer exposure. Their improvement was significantly affected when they were in a clinical group with a

Residential Iatrogenic Effects iv high proportion of externalizers. Under these conditions, they were the only group to show significant worsening over time. Gender differences and clinical implications are also examined and discussed.

Residential Iatrogenic Effects v Table of Contents Introduction……………………………………………………………………………………1 Method……………………………………………………………………………………….33 Results……………………………………………………………………………..…………46 Discussion……………………………………………………………………………………70 References……………………………………………………………………………………86 Appendices…………………………………………………………………………………...94

Residential Iatrogenic Effects vi List of Tables Table 1: Means and Standard Deviations for Narrow Band and Broad Band Scales on the Child Behavior Checklist by Parent Report ……………………….…………...…..36 Table 2: Means and Standard Deviations for Narrow Band and Broad Band Scales on the Teacher Report Form by Wediko Staff Report Averaged Over Three Assessments ………………………………………………………………39 Table 3: Behavior Group as Determined by Internalizing and Externalizing Behaviors on TRF…………………………………………………….…………………….40 Table 4: Peer Assessment Subscale Reliability and Intercorrelations…………….……..44 Table 5: CGI Means and Standard Deviations by Aggregated Staff Report…………….45 Table 6: Girls' Clinical Groups Externalizing Proportions………………………………53 Table 7: Boys’ Clinical Groups Externalizing Proportions……………………………...54

Residential Iatrogenic Effects vii List of Figures Figure 1. Means of antisocial influence for non-clinical, externalizers, internalizers, and mixed behavior children…………………………………….........................47 Figure 2. Boys’ change in antisocial influence by child and group type………………...55 Figure 3. Boys’ change in total problem behavior by child and group type……………...60 Figure 4. Boys’ change in externalizing by child and group type………………………...61 Figure 5. Boys’ change in aggression by child and group type…………………………..62 Figure 6. Global change as measured by the CGI for child type X behavior group for boys………………………………………………………………………………69

Residential Iatrogenic Effects viii List of Appendices Appendix A: Teacher Report Form (TRF)……………………………... …………………….94 Appendix B: Peer Assessment Interviews…………………………………………………......96 Appendix C: Clinical Global Impressions (CGI)………………….………….. ………………97 Appendix D: Parental Consent Form………………………………………………………...102

Residential Iatrogenic Effects 1 Youth in Residential Treatment: Deviant Behavior and Iatrogenic Effects in Peer Groups For years, people have aggregated youth in group treatment programs. In classrooms, after school programs, juvenile detention centers, and clinical settings, group treatment has been viewed as helpful, cost effective and efficient. Meanwhile, developmental researchers have been investigating how peers negatively influence one another in their day to day interactions outside of treatment and school groups. These observations of negative peer effects in naturally occurring peer groups have slowly raised concern about whether these processes may also occur when deviant youth are aggregated for treatment. Recently, greater attention has been paid to the ways in which peers influence each other’s behaviors in treatment contexts. Peer reinforcement can undermine adult guidance, with a rate of 9 to 1 reinforcing both positive and negative acts from peers compared to adult staff within institutional settings (Buehler, Patterson, & Furniss, 1966). Specifically, there has been concern over whether peers are capable of contributing to the development of deviant behaviors including rule-breaking and aggression (Dodge, Dishion, & Lansford, 2006). These concerns have been raised for school settings, community-based programs, and residential treatment programs, but little research has been done on the last of the three. Studies have shown that sometimes there are unintentional negative effects of bringing youth together for treatment because of peer contagion and direct reinforcement for deviant behavior by peers within a treatment group (Dodge et al. 2006). There have been many recent attempts to understand how peers influence each other, yet many questions still remain, particularly in the field of mental health and peer influence. The environment, personality characteristics of

Residential Iatrogenic Effects 2 children, age, gender, and many other variables must be examined to understand under what conditions peers are most likely to influence and be influenced by one another in treatment. This literature review will investigate what types of peer effects have been identified for different types of treatment groups, how peer influences can be helpful or harmful for groups, and under what conditions negative peer contagion effects are most likely to occur. More specifically, it will review what is already known about peer influence regarding aggression and aggregating aggressive youth in treatment, examine what remains to be known, evaluate whether peer contagion is harmful for youth in residential treatment in particular, and discuss what this means for mental health programs. Early Concern with Iatrogenic Effects Deviant youth, or those who engage in rule-breaking and aggressive behaviors, are often disruptive to groups of typically developing children, and often have needs that are difficult to address in mainstream settings (Dishion, Spracklen, Andrews, & Patterson, 1996). Logistically, it makes sense to remove these youth from such settings so as not to interrupt productive group processes (Dodge et al., 2006). Dodge et al. (2006) also argued that it is financially efficient to treat deviant youth in groups rather than individually, and that the general public prefer deviant youth be segregated so they cannot disturb mainstreamed children. However, research on treatment effects for aggressive and deviant youth has shown increasing concern about removing these youth from the influence of mainstream peers and aggregating them into deviant groups for education and treatment.

Residential Iatrogenic Effects 3 Ladd & Mize (1983) also suggested that aggregating children into group treatment is efficient and makes sense conceptually. Group therapy encourages opportunities for peer support and feedback. Similarly, group treatment is generally viewed as time- and cost-effective (Dennis, Godley, Diamond, Tims, Babor et al., 2004). On the contrary, Dodge et al. (2006) suggested that segregating deviant peers from their prosocial counterparts is, in fact, detrimental to youth who have problems with delinquency and aggression. Most programs in place for delinquent and aggressive youth, however, including alternative schools, juvenile justice facilities, and therapy groups do separate deviant youth from mainstream peers and aggregate them with one another (Dodge et al., 2006). In school settings, tracking students based on academic skill, emotional disturbances, and special education often leads to segregated peer groups. Rather than mainstreaming antisocial students in the larger classroom where they can learn from well-adjusted peers, these “special” children are pulled out into self-contained classrooms where there are no positive role models to demonstrate appropriate behaviors (Dodge et al., 2006, p. 7). Similarly, mental health settings generally group deviant children together. Halfway houses, group homes, and day treatment programs can be considered as forms of “batch processing” through which problems are dealt with as a group and peers are easily influenced by each other as it is difficult to find privacy and independence in such a setting (Dodge et al., 2006, p. 8). Despite these longstanding practices, it is only over the past decade or so that it has become apparent that grouping deviant youth may have harmful effects on treatment outcomes. Research in naturalistic settings has revealed some cause for concern regarding how and in what capacity peer contagion can occur. A large portion of the literature deals

Residential Iatrogenic Effects 4 with assigned groups which force the aggregation of deviant individuals, rather than naturally occurring groups, such as cliques and gangs, where youth are “natural,” not forced members (Klein, 2006). Snyder, Schrepferman, Oeser, Patterson, Stoolmiller et al. (2005) conducted observational research that suggested that selective relationships form among aggressive youth as young as preschool and elementary school. Aggressive kids actively selected each other as friends and tended to associate more with one another than with kids identified as normative (Snyder, West, Stockemer, Givens, & Almquist-Parks, 1996; Snyder et al., 2005). This self-selection into aggressive groups was evident even for young elementary children, and when these aggressive peer affiliations occurred, children were negatively affected by it. Snyder, Snyder, Schrepferman, & Stoolmiller (2003) suggested that although young children have limited access to peers, there are still plenty of opportunities for them to engage in deviant behavior in places like playgrounds and backyards, where there is little adult supervision. These behaviors then carry into the classroom. High levels of both overt and covert conduct problems predicted association with deviant peers in elementary classrooms. Although it seems relevant only outside of the classroom, children’s choice of peer affiliations and sensitivity to peer influence impacted their classroom adjustment (Snyder et al., 2005). Just as selective affiliation and its negative consequences are evident among aggressive preschoolers (Snyder et al., 1996), research by Duncan, Boisjoly, Jremer, Levy, & Eccles (2005) indicated that firstyear, male, college students randomly assigned to live with peers with a history of alcohol use were four times more likely to increase drinking over the first 2 years of college. Duncan’s research suggests that natural environments can reinforce peer contagion effects of aggressive and/or maladaptive behaviors.

Residential Iatrogenic Effects 5 With this developmental research as the foundation, there has been an explosion in research on peer contagion among aggressive and delinquent youth in group treatment over the past 10 years. Dishion, McCord, and Poulin (1999) are pioneers in this research. They were the first to draw attention to possible “unintended negative consequences,” (Phillips, Smith, Dumas, & Prinz as cited in Dodge et al., 2006, p. 297) of group treatment, and hypothesized that the presence of deviant youth in a peer group may offset the positive gains of an intervention program. Dishion et al. described the Adolescent Transitions Program (ATP) (Dishion, McCord, & Poulin, 1999), an intervention program for antisocial youth involving both peer and parent treatment components. High risk boys and girls, as defined by parent and teacher reported levels of rule-breaking behavior, were randomly assigned to one of four intervention conditions including a (1) parent focus only, (2) teen focus only, (3) both parent and teen focus, and (4) a “self-directed change” placebo group. Based on developmental models of aggression and conduct disorder, the researchers hypothesized that the combined teen and parent group would be most effective in lessening problem behavior (Dishion & Andrews, 1995). Short-term results were mostly positive, particularly regarding family interactions, family tension and conflict. Long-term analyses, however, yielded increases in externalizing behavior for the teen-focus group compared to the control conditions. Follow-up tests showed iatrogenic effects, or unintended negative effects of treatment, within the teen focus group for delinquency and substance use. Participants who were randomly assigned to the peerfocus intervention showed an increase in tobacco use and teacher reports of delinquent behavior beyond the levels reported for youth who had received only minimal intervention at the 3-year follow up (Poulin, Dishion, & Burraston, 2001). The oldest

Residential Iatrogenic Effects 6 group of teens (postpubertal aged 15-16) with the highest initial level of problem behavior was most susceptible to iatrogenic effects. Dishion et al. (1999) reasoned that the older youth were more affected because older teens tended to “mobilize more group attention” than younger deviant youth, thereby encouraging more “deviancy training,” a process variable that helped explain this unintended negative treatment effect. Dishion et al. (1999) define deviancy training as the “process of contingent positive reactions to rule-breaking discussions.” Deviancy training progresses from positive peer reactions to deviant talk to positive reactions to engagement in deviant behaviors. Youth are reinforced through laughter, attention, and interest of other peers. If they misbehave and peers laugh or joke about it, the child is likely to engage in the attention-seeking behavior again, even if it is negative. The ATP findings suggest that interventions targeting certain types of high-risk youth need to have a family focus that encourages caregivers to structure environments and limit their child’s time with deviant youth to avoid inadvertent deviancy training (p. 762). Incorporating parents in interventions for high-risk youth has shown some positive effects regarding diminishing delinquent behavior and iatrogenic effects. Perhaps one of the earliest studies to raise concerns about deviant youth aggregation dates back to 1939. The Cambridge-Somerville Youth Study (CSYS) collected a sample of approximately 500 children between the ages of 5 and 13. Boys were identified as “troublemakers” or average based on community referrals. The boys were randomly assigned to control and treatment groups and used a paired-samples design to compare changes in deviant behavior. The Cambridge-Somerville intervention program included group activities for the boys including after-school games, as well as

Residential Iatrogenic Effects 7 individual aspects, like tutoring. Initially, the pairs of boys were very similar in delinquent behavior, intelligence, age, neighborhood, home stability, aggression, and other variables. The treatment group received visits on average twice monthly by counselors in their homes who encouraged them to participate in community activities, sports programs, and local jobs. McCord (1978) conducted a 30 year follow-up review of boys in the program. She concluded that boys who had been in treatment groups were more likely in adulthood to have become repeated offenders for various crimes compared to boys who had not been involved in the intervention program. Naturally, the question arose of how treatment could possibly have led to such negative effects for young boys. Later research by McCord (1992) on the same sample yielded interesting answers to this question. She found that boys who were assigned to a summer camp as a form of intervention, and who had spent two consecutive summers there showed the most adverse outcomes, and they were most likely to become delinquent. Results suggested undesirable outcomes were more likely for youth who were in intense, long-term treatment in which they were aggregated with other delinquent youth. Outcomes were undesirable if the participant (a) died prior to reaching the age of 35, (b) was convicted for a serious crime, (c) was diagnosed as an alcoholic or labeled as psychiatrically impaired (schizophrenic, manic depressive, etc). There were no main effects for age at which treatment was started. Both CSYS and the Adolescent Transitions Project (Dishion et al., 1999) showed that aggregating peers with behavior problems can be associated with increases in negative behaviors including delinquency and substance use. Dishion suggests that just as deviancy training occurs in naturalistic settings, it may also occur in treatment programs.

Residential Iatrogenic Effects 8 With these concerns about possible deviancy training in mind, Cho, Hallfors, and Sanchez (2005) examined Reconnecting Youth, an after school, non-academic program for teens. The program aimed to reconnect underachieving high school students to their families, schools, and communities through courses in self esteem, communication, and decision making. Goals of the program included decreasing drug involvement, school problems, and increasing GPA and mood management. Researchers clustered 1218 fifteen year-olds into highly structured, small peer groups of 10-12 high-risk students from two diverse urban school districts, led by a trained teacher. Comparing grade point average, anger, school connectedness, conventional peer bonding, and peer high-risk behaviors pre and post intervention showed mixed results at the end of the semester. For example, compared to the control group, the experimental group who participated in the program showed significantly decreased levels of delinquency as measured by a scale including such items as “In the last 30 days, how many times did you… (a) get into a physical fight with someone? (b) take life-threatening risks? (c) shop-lift, steal, or ruin someone’s property?” There was also a positive post-intervention effect for smoking in the experimental group. Youth in the experimental group did, however, show negative outcomes for conventional peer bonding compared to the control group, where youth in the experimental group fostered showed less conventional peer bonding. That scale included items such as, “How many of your close friends… (a) attend church or religious group activities? (b) are active in school/community sports, clubs, or activities? (c) plan to go to college? (d) spend lots of time with their families? (e) know your parent(s) well?” Despite mixed outcomes at immediate post-intervention, there were only negative effects at a six-month follow-up, including less optimal scores on measures of grade point

Residential Iatrogenic Effects 9 average, anger, school connectedness, conventional peer bonding, and peer high-risk behaviors. Cho et al. speculated that the aggregation of high-risk youth kept them from fostering friendships with prosocial peers, and instead created more opportunities for friendships with deviant peers. To more closely examine the process of deviancy training that may have given rise to the iatrogenic effects observed by McCord (1992), Dishion, Spracklen, Andrews, and Patterson (1996) recorded conversations between adolescent dyads and documented levels of normative versus rule-breaking conversations, as well as reinforcing behaviors such as laughing or pausing in response to deviant talk. Dyads were identified as delinquent if both boys had been arrested, mixed if only one boy had been arrested, and non-delinquent if neither boy had been arrested. Delinquent dyads showed a linear relationship between reinforcing rule-breaking conversation and laughter. In mixed and non-delinquent dyads, there was reinforcement for normative talk. Furthermore, followup longitudinal analyses showed that deviancy training predicted future delinquent behavior. Other early research on iatrogenic effects has yielded results similar to Dishion et al. (1996; Dishion et al., 1999) and has pointed to possible moderators of the processes Dishion describes and the iatrogenic effects people have increasingly observed. Feldman’s (1992) St. Louis Experiment assigned 283 early adolescent high-risk boys and 438 low-risk boys randomly into three types of groups of about 10-15 boys: entirely highrisk youth groups, entirely non-referred youth groups, and mixed groups that included mostly low-risk boys with one or two high-risk boys. Furthermore, the groups either had an experienced or inexperienced leader who provided three different levels of

Residential Iatrogenic Effects 10 intervention within groups. The “traditional” groups were loosely structured and focused on group discussion. Behavioral groups were more structured, but focused on group interventions. The minimal intervention groups provided less structure and less therapeutic tasks for the groups to complete. Feldman’s study aimed to explain how deviant youth behave when mainstreamed in a group of predominantly low-risk youth, as well as how low-risk youth react to the presence of their high-risk peers. Using overall change as reported by therapists, direct observation, and selfreports, Feldman measured differences between first and last group sessions, and results suggested significant differences between group conditions. High-risk boys assigned to homogeneously high-risk groups showed increased problem behaviors, while high-risk boys assigned to mixed (mostly low-risk) groups showed decreased problem behaviors. High-risk boys in the all-deviant groups with an inexperienced leader showed the most significant increases in antisocial behavior. Feldman (1992) points out that if high-risk youth must be aggregated in deviant youth groups, there are ways to control iatrogenic effects, including placing high-risk youth with experienced group leaders and in highly structured programs. High-risk boys placed in the behavioral intervention group that focused on high levels of structure and group cooperation showed improvement in both all-deviant groups and mixed groups. Furthermore, Feldman suggests that mixing high and low-risk youth is beneficial in that there are very few negative effects for the low-risk youth, while increasing prosocial behavior of high-risk youth. Factors that Moderate the Risk of Iatrogenic Effects in Group Treatment Although there is clear evidence of the potential for iatrogenic effects, or at least dampened treatment effects when aggressive and deviant youth are aggregated in group

Residential Iatrogenic Effects 11 treatment, it is important to note the conditions that have been shown to mute or enhance these effects. For example, Dishion et al.’s ATP (1999) showed that pre-pubertal adolescents (11-13 years) were less vulnerable to iatrogenic effects than adolescents involved in the program (15-16 years), suggesting that perhaps developmental stage is an important determinant of whether group treatment will have iatrogenic effects. Similarly, structure seems to matter. Intervention programs with trained leaders were better able to control the group’s delinquent behaviors (Feldman et al., 1983). Feldman et al.’s research suggested that the expertise of the group leader was a protective factor for adverse peer effects, where youth randomly assigned to a group with a more experienced leader were less influenced by the aggregating of deviant peers. So far, the literature on whether group treatment is harmful has not provided a definite answer. One factor to consider is when the effects start to show. In the short run, group treatment might show many positive outcomes. Long term effects and the processes that mediate them, however, could be a cause for concern. For example, group treatment might introduce youth to delinquent friends they can easily socialize with after school, before or after treatment, and during other unsupervised free time. As these deviant youth spend more time together outside of treatment, they may run the risk of long-term negative consequences. Another question that remains regarding whether group treatment is harmful is how the presence of well-adjusted youth affects the process. These are probably only a few of the many factors that might influence how and when deviant youth are negatively influenced by peers in treatment. It seems in the short-run that some treatment, even if it aggregates externalizing youth, is better than no treatment at all (Ang & Hughes, 2002). A meta-analysis of 38

Residential Iatrogenic Effects 12 studies of primarily externalizing youth ages 6 through 18 years who received some form of social skills training outperformed 73% of externalizing children receiving no treatment or the control treatment in immediate post-treatment assessments. One important shortcoming, however, was that a majority (51.4%) of the studies examined failed to report follow-up data, so it is difficult to determine from this meta-analysis whether there are harmful follow-up effects of aggregated group treatment. Just as Cho et al. (2005) found negative outcomes at follow-up tests, Dishion et al. (1995) found evidence for adverse effects of aggregating at-risk teens only at post-treatment. Furthermore, 7 out of 8 negative effect sizes (Ang & Hughes, 2002) revealed in the metaanalysis came from homogeneous groups, which suggests that aggregating antisocial youth for teaching social-cognitive and behavioral skills without including any nonantisocial youth might yield iatrogenic effects and might dampen treatment effects. Bronfenbrenner (1979) proposed an ecological framework to explain the different levels of influence regarding individual differences in social development of children and adolescents. With regard to peer contagion and iatrogenic effects, one could extend levels of influence to include the level of the individual, the program or intervention service level, and the context level in which the program or intervention is fixed. Characteristics of the Individual. Brofenbrenner’s ecological framework suggests that the characteristics of a child as a recipient of potential peer influence may be important to examine as we seek to explain who is most vulnerable to negative iatrogenic effects of group treatment (1979). Age, gender, behavioral history, and temperament are some factors that may influence the way a child understands and responds to an intervention, and to peer influence within treatment intervention (Dodge et al., 2006).

Residential Iatrogenic Effects 13 Poulin, Dishion, and Haas (1999) examined the “Peer Influence Paradox.” The researchers examined the reciprocal relationship between delinquent behavior and poor peer relations. Previous research by Dishion, Andrews, et al. (1995) suggested that the quality of friendships of antisocial children were generally shorter in duration, characterized by bossiness, and were generally less satisfying. Poulin et al. compared quality of friendship, as measured by report of positive and negative aspects of the friendship as well as reciprocal liking, for 13-14 year old boys identified as highly antisocial at age 9-10 compared to boys identified as normative at the same age. In addition to child-report of the friendship, each child was asked to pick the peer “with whom he spent the most time,” and interviewers rated their impressions of relationship quality in dyadic observations. Compared to normative boys, antisocial boys showed almost no agreement in friendship perceptions. Dishion, Andrews, et al. (1995) suggested that these types of friendships often develop out of convenience, rather than common activities and hobbies. Rather than actively selecting friends they like, antisocial boys more often associated with other available peers in their neighborhood or other ostracized peers at school in lower quality friendships than normative boys of the same age. Although the authors hypothesized that higher quality friendships were more reinforcing, and therefore led to greater tendency for peer influence, Poulin et al. (1999) found the opposite. Boys with very low quality friendships who were delinquent at age 13-14 were the ones who showed the most significant increases in delinquent behavior in response to peer reinforcement. Dishion (2000) examined cross-setting consistency in deviant peer groups and elevated deviancy training. In a sample of 224 high-risk adolescent (average age = 12

Residential Iatrogenic Effects 14 years) boys and girls, Dishion found that externalizing youth who rated high on both internalizing and externalizing scales (comorbid) were more prone to deviant talk than youth who showed only internalizing symptoms and children who showed normal levels of both internalizing and externalizing behaviors. Externalizing youth were defined as children whose problem behavior included symptoms of oppositional defiant disorder and conduct disorder (including rule-breaking and aggressive behaviors. Internalizing youth were those who scored high on anxiety, depression, and somatic symptoms. Mixed internalizing/externalizing youth showed the greatest vulnerabilities for negative effects of deviancy training in peer groups (Dishion, 2000). Dishion suggested that comorbid youth tend to experience negative life events early on in childhood, which lead to emotional distress and problem behavior in school. School problems lead to depression and peer rejection, which ultimately lead to involvement in deviant peer groups. Because deviant peer groups support problem behavior, Dishion argues, these groups lead to opportunities for deviant talk, which leads to the reinforcement of deviant talk, and ultimately more delinquent behavior. Dishion’s hypothesis complements his theory that externalizing youth are the ones doing the deviant talk. He suggests that mixed youth are influenced by the externalizing deviant talkers. Features of the Group and Program Type. The literature suggests that programs and policies that group deviant peers repeatedly yield harmful effects. While many researchers have expressed concern that deviant youth can negatively impact peers, Dishion and Dodge (2005) warn specifically against aggregating externalizing youth. They authors suggest that homogeneous, externalizing classrooms and peer groups in community-based or juvenile detention centers can amplify problem behavior. So what

Residential Iatrogenic Effects 15 does this mean for mental health treatment, particularly residential mental health facilities, where it is often more practical, cost-efficient, and effective to utilize group therapy? Dishion, Dodge, & Lansford (2006) offer insight into what types of programs and settings are mostly likely to give rise to problems with peer contagion and deviancy training: (1) Any group therapy in which the ratio of deviant to nondeviant youth is high, (2) Group therapies with poorly trained leaders and lack of supervision, (3) Group therapies offering opportunities for unstructured time with deviant peers, and (4) Group homes or residential facilities. Contrary to much of the literature on aggregating deviant youth, Mager, Milich, Harris, and Howard (2005) examined the effects of mixed groups comprised of both high- and low-risk youth, and demonstrate how group composition can influence outcomes. With a sample of 129 6th and 7th graders who had been referred based on teacher-reported ratings of conduct problems in the classroom, the researchers developed a program aimed to help youth develop different skills. The focus of the program was the development of problem solving skills for social groups including teasing, peer pressure, and disagreements between peers. Measures of psychosocial functioning were completed at least 3 months prior to the start of the program, 3-5 months into the program, 1 month after the program had finished, and 6 months after the program. Measures used included the Child Behavior Checklist (CBCL), Attitudes about Alcohol and Drugs, Attitudes Toward Delinquency, Social Skills Rating System, School Adjustment Scale, Peer Association (sociometric measure), and behavioral observations (the number of times a child violated the rules).

Residential Iatrogenic Effects 16 Mager et al. (2005) found that students in the pure externalizing group decreased externalizing behaviors and increased prosocial behavior in comparison to high-risk youth in the mixed group condition. Furthermore, there was less deviant talk in the pure externalizing group compared to the mixed group. In this study, aggregation of aggressive youth led to positive results, while the mixed groups showed less positive outcomes. Furthermore, Mager et al. (2005) suggest that perhaps group size matters. In their research, mixed groups had a larger average size, so perhaps smaller groups are more effective. To further explain the observed effect, Mager described disengagement of at-risk members of mixed groups when they are a minority of the group. The group leaders tended to focus on the prosocial majority, rather than the at-risk deviant minority. Individual and Group Fit Bronfenbrenner’s framework explains the separate contexts that can influence peer contagion effects including the individual, context and environment, and interactions between the individual and environment (as cited in Dodge et al., 2006). Perhaps the most important aspect of that framework is the consideration of interaction effects. That is, no one aspect of the framework is likely to completely explain peer contagion effects, rather interactions of the several dimensions will be essential to consider. Several researchers have examined how individual and group factors interact to predict social experiences and outcomes. This work bears both indirectly and directly on the issue of iatrogenic effects of aggregating antisocial youth in treatment. Hanish, Martin, Fabes, Leonard, and Herzog (2004) explored the concept of homophily, the selective affiliation that tends to occur when aggressive peers are more likely to interact with other externalizing peers than any other children (Cairns, Cairns,

Residential Iatrogenic Effects 17 Neckerman, Gest, & Gariepy, 1988; Espelage, Holt & Henkel, 2003). The homophily hypothesis also suggests that youth who actively seek out similar peers are more influenced by these peers because they model and reinforce behavior within the group and extinguish any behaviors that are not reinforced. In a study of middle school students (6th through 8th grade), boys and girls reported bullying and fighting frequency within their school by self-report questionnaire (Espelage, Holt, & Henkel, 2003). Students also completed peer-nomination tasks in which they reported the names of students who often bullied peers, as well as a list of their own close friends in the class. Results indicated that homophily existed within peer groups, in that students tended to cluster with peers who fight, bully, and tease at a similar frequency. Wright, Giammarino, and Parad (1986) studied peer group and social status of 138 high-risk, emotionally and behaviorally disturbed youth in a short-term residential treatment program. Wright et al. hypothesized that person-group similarity is a predictive factor of child status within peer group. The researchers suggest by that hypothesis that an aggressive child may be viewed as unpopular in a group of prosocial youth. When that same aggressive child is placed in a peer group comprised of other aggressive youth, however, the target child’s aggressive behavior is viewed as less deviant, and his social status may increase. This can lead to reinforcement of the aggressive behavior. Findings confirmed this person-group effect on social behavior and social status within treatment groups. Another aspect of individual-group fit to consider is the baseline level of aggression. (Boxer, Guerra, Huessman, & Morales, 2005). High-risk 3rd through 6th grade

Residential Iatrogenic Effects 18 boys and girls as reported by teacher-assessed classroom behaviors were compared on individual levels of aggression to the mean aggression of the group. Children who demonstrated below-average aggression became more aggressive, while children who were higher in aggression compared to the group mean decreased in aggressiveness. Boxer et al. suggest that baseline effects are greater when the target child’s aggression is farther from the mean aggression of the group. That is, individual changes in aggression were related to the relative aggression of the group, where children were pulled toward the group mean. That was true for both highly aggressive children in less aggressive groups, and less aggressive children in highly aggressive groups. The researchers referred to this tendency as the “discrepancy-proportional peer-influence effect,” suggesting that iatrogenic effects exist for children in aggregated groups, especially when they deviate the most from group norms. Boxer et al.’s findings, however, are a bit general. For instance, what happens to the very low aggressive child in a moderately aggressive group? Similarly, what about the highly aggressive kid in a moderately aggressive group? Boxer et al.’s results only describe extreme group conditions, and do not account for moderate group levels of aggression. Despite these lingering questions, Boxer et al. (2005) demonstrated that aggregation and the proportion of deviant youth within the group could serve to either increase or decrease problem behaviors. Some research has demonstrated iatrogenic effects for aggregating deviant youth (Feldman, 1992), while others have not shown such effects (Mager, 2005). Similarly, some studies have shown dampened treatment effects (Boxer et al., 2005). The previous research suggests that multiple factors contribute to the occurrence of iatrogenic effects, including the group composition, proportion of deviant talkers, group levels of

Residential Iatrogenic Effects 19 aggression, and child-group fit. Much of the research indicates that gender differences may exist, yet most of the research focused only on males. Clear potential for iatrogenic effects has been shown in outpatient school-based and correctional settings, but other treatment settings, most notably residential mental health treatment settings, have not been studied. Efficacy and Effectiveness of Residential Treatment Residential treatment is a much utilized part of the spectrum of services for youth with serious emotional and behavioral problems, but given the recent attention to negative peer influences in group-based treatment there is controversy over whether it is an effective option for aggressive and delinquent youth. Dodge, Dishion, and Lansford (2006) strongly advise against residential treatment for fear of iatrogenic effects, but in actuality, there has been very little research done on negative peer influences and iatrogenic effects in residential treatment. Residential treatment does aggregate youth with aggressive behaviors, which has potential for harmful iatrogenic effects. However, it also has many of the features that are known to mitigate against iatrogenic effects including a high staff:child ratio, highly trained staff, highly structured activities and days, and minimal unsupervised time. Hair (2003) explains that residential treatment provides a consistent, nurturing environment with predictable, consistent expectations. These routines and patterns help shape behaviors and emotional responses. Research contends that grouping aggressive and delinquent youth yields iatrogenic effects because those individuals feed off each other’s antisocial behaviors and escalate aggressive situations. With these controls in place, however, iatrogenic effects may not occur or may not be as strong as researchers suspect.

Residential Iatrogenic Effects 20 Although there has been little research on negative peer influence in residential treatment, there is a body of research on the effectiveness of residential treatment. Generally, positive behavior effects are shown over treatment (Frensch & Cameron, 2002; Connor et al., 2002), although this varies some over studies. Much of the literature indicates improvements during the course of and by the conclusion of residential treatment including less pathology at discharge (Connor et al., 2002), and reduction of high risk behaviors, aggression, depression, and psychotic features (Lyons Terry, Martinovich, Peterson, & Bouska., 2001). McCurdy and McIntyre (2004) outline factors that are particularly beneficial in residential treatment. They explain that while there have been many changes made in quality and availability of community-based programs, there has been little change in residential treatment over the years. Particularly with admissions to residential treatment centers on the rise, McCurdy and McIntyre point out the need to update practices in their programs and recommend the stop-gap model of residential treatment. The stop-gap model emphasizes three levels of intervention including environment-based, intensive, and discharge-related interventions. Environment-based interventions are aimed at creating an environment that helps the individual move toward community-based treatment, as well as skill-teaching exposure including academic, social, problem-solving, and anger management skills training. Environment-based interventions should also employ a token economy that rewards individuals for positive, prosocial behaviors. McCurdy and McIntyre suggest that intensive interventions should include a functional behavioral assessment and behavior support plan catered to each individual child. Discharge-related interventions are aimed at preparing the individual to return to his

Residential Iatrogenic Effects 21 family and maintain the gains made through residential treatment. The stop-gap model suggests a comprehensive approach to residential treatment compared to previous models that were driven heavily by medical and psychoanalytic models. McCurdy and McIntyre (2004) suggest that the stop-gap model is beneficial because of its short-term, focused treatment. Furthermore, because the service is aimed at short term residency, youth spend less time fraternizing with antisocial peers, and the chances for deviancy training to occur are limited. Furthermore, the components of the stop-gap model are clear and specific, so staff can be well-versed in many facets of the program, rather than specialized in only one area. Similarly, short-term residential treatment may have lower staff turnover rates. In many long-term residential programs, a few, highly trained professionals are responsible for treatment. The ability to share treatment responsibilities among equally trained pre-professionals, eager to gain experience in the field may encourage staff to stay longer. Lastly, because the stop-gap model proposes shorter periods of stay, the overall cost of residential treatment will be reduced. Researchers are recognizing that shorter treatment programs are better. Shapiro et al. (1999) found in a long-term residential treatment program that almost all behavioral and emotional improvements took place within the first three to six months of the program. The present research examines peer effects within a short-term residential treatment program like the stop-gap model suggests. Despite its evidence for short term efficacy, one problem with residential treatment research is that it is often difficult to gauge how effective treatments are over the long run, as many studies measuring change limit their assessment to within-program behaviors. Frensch and Cameron (2002) explained that over all outcome studies of

Residential Iatrogenic Effects 22 residential treatment, there have been mixed results. Further attempts to identify factors that predict outcomes have met with limited success. Another very common issue with residential treatment seems to be transfer of gains. Many studies show fading treatment effects over time, and in some cases no prepost evidence of treatment effects when effects are measured several months after treatment has ended. Frensch and Cameron (2002) reviewed the effectiveness of residential mental health treatment programs for troubled youth, paying particular attention to the long- and short-term effects of treatment. Previous research suggests that long-term effects require extensive family involvement and encouragement. One continually positive predictor of long-term gains is family involvement. It is important to acknowledge, however, that many children who attend residential treatment programs have strained relationships with their parents, or are part of a dysfunctional family and may not receive the encouragement and attention required to make successful gains at home. An important factor that may help ease the transition from a residential program to home life is maintaining contact and support from families while the child is part of the program. Frensch and Cameron (2002) note that children who are discharged into stable, positive, and supportive environments show the most improved posttreatment functioning, yet they argue that not enough is known about residential treatment and why it has limited impact outside of the treatment program. Studies with positive long term effects show that transfer of gains is facilitated when interventions also target families and do careful aftercare planning. Hair (2003) examined what factors predict successful transitions for emotionally and behaviorally disturbed youth out of residential programs and back into families. She identified three

Residential Iatrogenic Effects 23 key factors including family involvement, stability of environment, and aftercare support. Knorth, Klomp, Van der Bergh, and Noom (2007) similarly suggested that parental involvement is a critical factor in predicting outcomes of residential treatment. Despite the difficulty demonstrating strong treatment effects outside of the treatment environment (especially as the follow-up period lengthens), to our knowledge, no research has shown the iatrogenic effects that Dodge et al. (2006) warn of when they recommend that residential treatment not be used for aggressive and delinquent youth. Iatrogenic Effects in Residential Treatment Programs. Previous research on possible iatrogenic effects in residential treatment at a summer residential treatment program run by Wediko Children’s Services (Cardoos, Zakriski, Wright, and Parad, 2008) addressed the question of unintended harm, with particular emphasis on the much studied process variable, deviant talk. Cardoos et al. investigated whether youth in residential treatment engage in deviant talk, what other behaviors characterize youth who engage in deviant talk, and whether it was related to negative outcomes. Although there is very little research on residential treatment, the existing literature suggests that youth in residential treatment programs, would engage in deviancy training, just as they do in most group treatment programs (Dodge et al., 2006). Cardoos et al. (2008) found that deviancy training did occur at Wediko, as evidenced by youth engagement in deviant talk (talking about breaking the rules, laughing when others broke the rules). Deviant talk increased over time for preadolescent groups, and most often was done by externalizing youth. Interestingly, deviant talk was associated with peer rejection, not peer support (as might be evidenced by peer acceptance). Indeed, children who engaged in deviant talk experienced less positive interactions with other children and with staff. Interestingly,

Residential Iatrogenic Effects 24 preliminary findings did not show evidence of iatrogenic effects. Instead, “deviant talk at the beginning of the summer was significantly related to improvement in aggression, rule-breaking, externalizing, and total behavior problems on the TRF” (p.5). Cardoos et al. then investigated whether the impact of deviant talk was moderated by the behaviors of the surrounding peer group. They found evidence of a person-group fit effect for deviant talk, where youth who engaged in deviant talk experienced dampened treatment effects when they were in a group that engaged in high levels of deviant talk compared to deviant talkers in low deviant talk groups. These youth did not worsen, but rather they improved less than deviant talkers who were in peer groups that were less receptive and encouraging of their deviant talk (for overall behavior change, aggression, and relations with adults). Cardoos, Zakriski, Wright, and Parad (2008b) did note one narrow iatrogenic effect for preadolescents, who showed increased aggression in response to peer provocation. These youth did not, however, show increased aggression in other contexts (e.g. to adult instruction or peer prosocial initiatives). This suggests that deviant talk to a receptive audience might be more likely to escalate aggression in negative peer interactions, particularly in preadolescent peer groups. But, it does not seem to increase aggression across the board. Cardoos et al. (2008) suggest that perhaps the harm aggressive and delinquent youth might cause in residential treatment is exaggerated. On the other hand, Cardoos et al. (2008b) acknowledge that peer acceptance and encouragement of aggressive and delinquent behaviors is possible even in a highly structured program like Wediko. More importantly, however, they explain that these negative effects can be controlled with constant supervision and monitoring, as well as individual behavior management techniques. Also of note is the suggestion for

Residential Iatrogenic Effects 25 heterogeneous peer groups within a clinical, at-risk population, as treatment dampening effects were most prevalent in more homogeneous groups. Cardoos et al. (2008) also describe a prior investigation showing that gender may be an important variable to consider. This study (Cardoos, 2006) examined the relationship between peer social status and aggression, as well as the treatment outcomes for aggressive versus non-aggressive girls and boys. She found that girls experienced more positive social reinforcement for rule-breaking behaviors in residential treatment. Specifically, aggression was positively related to social preference nominations from peers for girls, but not for boys. Aggressive girls, who received such peer support, were less likely to improve compared to non-aggressive girls, while the reverse was true for boys, who received little peer support for deviant behavior. Interestingly, boys low in deviant and delinquent behavior were most vulnerable to peer contagion effects and showed the least overall improvement from the treatment program. Dodge et al. (2006) mention almost nothing about gender differences in delinquency and deviant talk, which means either that there are no effects for gender, or more research is necessary. Based on Cardoos et al.’s results, however, we know that the latter must be true. It is possible that there are gender differences in the processes of aggressive behavior and reinforcement. It will be important to study these gender differences. Other related research at Wediko by Kretsch, Olds, Wright, Zakriski, and Cardoos (2008) investigated whether levels of antisocial behavior (for individual and groups) might help clarify peer group experiences and possible iatrogenic effects within residential treatment. They focused on child-group fit is as a predictor of social experiences in short-term residential treatment, but focused on aggression rather than

Residential Iatrogenic Effects 26 deviant talk (as in Cardoos et al. 2008). Kretsch et al. expected that child-group similarity would explain the types of social events children experienced, and the way in which they responded to these events. Furthermore, Kretsch et al. (2008) hypothesized that children who were “social misfits,” or children whose behaviors were not similar to that of their group would experience fewer positive interactions with peers, and be less likely to respond in a positive ways to peer attention. To determine whether there was a goodness of fit between child and group, the individual child’s aggression level was calculated from extensive observations of overall aggression rates, as well as the average aggression level of the group (minus the target child’s score). High aggressive children in a mismatched low aggressive group experienced a notably low frequency of positive events. These children also showed lower rates of positive reactions to both neutral and aversive events. While person-group similarity had little effect on overall behavior change, “misfit” children showed the most dramatic change over the summer, compared to children whose behaviors matched those of the group. Kretsch et al.’s results showed that misfit non-delinquent children placed in aggressive groups had little significant change in aggressive behavior, and instead increased their prosocial behavior. Their findings contradict the deviancy training hypothesis, which suggests that non-delinquent children placed in an aggressive peer group will show increased levels of aggression by peer contagion. Much was learned from these initial investigations of negative peer influence in residential treatment. Specifically, deviant talk exists in residential treatment settings, but there is little evidence of harm from either deviant talk or from being in treatment with aggressive youth. For deviant talk, there appears to be a synergistic effect between person

Residential Iatrogenic Effects 27 and group, where engaging in the behavior around others who engage in the behavior can dampen treatment effects (and in some cases show narrow evidence of worsening). For aggression, there is more of a misfit effect that influences social experiences over the course of treatment. Gender appears to be important, but very little is known. Age may also influence peer group effects in residential treatment. It may be that adolescents are more influenced by peers, while younger children are immune to these contagion effects. Clearly, more work is needed to see who is most at risk for engaging in negative peer influence in residential treatment, who is most at risk of harm from it, and what peer group processes might be especially problematic. The Present Study As mentioned above, residential treatment has many of the features that previous research suggests would be helpful for reducing negative peer influences and risk for iatrogenic effects. Like many residential treatment programs, the Wediko summer treatment program is a highly structured, behavioral program with minimal unstructured time. Children are constantly monitored by highly-trained staff, who engage them in ageappropriate, stimulating activities from wake-up routines to bedtime daily. Clinical groups are comprised of same-age, same-sex children, and older and younger kids are not mixed. Also, the groups are behaviorally heterogeneous, which has been linked to low risk for iatrogenic effects in different types of settings (Cardoos et al., 2008). Despite the beneficial characteristics of residential treatment, there remains controversy over whether it is an effective approach to mental health treatment for aggressive children. While there has been extensive research on the aggregation of deviant and aggressive youth in the classroom, juvenile detention centers, after-school

Residential Iatrogenic Effects 28 programs, and other types of treatment, little attention has been paid to aggregation of aggressive youth in residential mental health treatment. Residential treatment effectiveness research has shown, at the very least, short-term positive effects for most children, so where is the harm? Additionally, recent research probing for negative peer influences within the Wediko residential treatment setting has found evidence of treatment dampening under certain conditions, but very little evidence of harm. The present study took a step back from previous research in this setting that specifically sought to describe deviant talk and its effects in residential treatment. In doing so, this research examined more broadly the ways aggressive children influence peers in residential treatment. Specifically, the present study examined whether residential treatment is problematic for aggressive youth and/or for those in treatment with them. Special attention was paid to the differences between purely externalizing aggressive youth and mixed externalizing/internalizing youth. There has been little research on mixed youth, and no research has been done on mixed youth in residential treatment outside of Wediko. Both the interactions these types of children have with their non-aggressive and internalizing peers, and how these interactions influence the effectiveness of the treatment program were examined. Very little research has paid attention to the important distinction between purely externalizing youth and youth who display both externalizing and internalizing behavior problems. The latter group is wellrepresented in mental health settings, including residential treatment. Thus any effort to understand the risk of deviancy training or iatrogenic effects in residential treatment must investigate possible differences between these groups in how they influence or are influenced by the peer group. Dodge et al. (2006) suggests that the high rate of

Residential Iatrogenic Effects 29 comorbidity in youth in mental health settings presents challenges regarding residential management and community safety. In turn, “the added risk that comes with aggregating these severely deviant youth in groups may be small in contrast to these other challenges” (p. 373), so additional harm from being aggregated is unlikely. It is important to note that there are more comorbid than pure externalizing youth at Wediko because it is a mental health setting, and severely externalizing children without comorbidity are more often referred to juvenile detention centers or other types of correctional treatment programs. It is in residential correction facilities where research has demonstrated a strong need for caution (Dodge et al., 2006). It was expected that both pure externalizing and mixed internalizing/externalizing youth would contribute to negative behaviors and influence in the clinical peer groups at Wediko. However, because seminal findings on deviancy training and iatrogenic effects were noted for delinquent (often pure externalizing) youth (Dishion et al., 1999), we expected externalizing children to be the most problematic by engaging in deviancy training. To test this, externalizing and mixed children were compared on behaviors in the peer groups including antisocial influence (deviancy training behaviors), prosocial influence (promotion of prosocial behaviors), antisocial behavior, and prosocial behavior. The present study then examined the effects of having too many of these pure externalizing children together in one group. Past research at Wediko and elsewhere suggests that heterogeneous groups may help reduce the risk of iatrogenic effects, whereas homogeneous groups can increase risk. Using the natural variability that exists in treatment group composition at Wediko, we formed high externalizing and low externalizing groups, and investigated the many possible negative influences a cluster of

Residential Iatrogenic Effects 30 externalizing children can have on the group. For example, does the aggregation of externalizing children in a group lead to the escalation of their negative influence? Are the negative influences of externalizing youth contagious to mixed internalizing/externalizing children? Furthermore, what about the non-aggressive children? Previous research regarding negative peer influences suggests that non-aggressive youth are unharmed by the strategy of mixed aggressive and non-aggressive groups (Dishion et al., 1996), but that is based on research with non-symptomatic, non-aggressive children. The Wediko sample, however, includes symptomatic, non-aggressive children. It was expected that non-aggressive children in high externalizing groups would show lower or negative treatment effects than non-aggressive children in low externalizing groups because of their clinical vulnerability. These non-aggressive children may see that staff pays more attention to bad behaviors and act out to get similar attention. Furthermore, the nonaggressive children may feel left out or less connected to the peer group if they don’t show externalizing behaviors. Adopting those negative behaviors could result in lower or negative treatment effects for these children compared to non-aggressive children in nonexternalizing groups who would already feel comfortable and not have to compete as much for staff attention. Also of interest was the expectation that externalizing children, especially when surrounded by many externalizing peers, would show less positive treatment effects over the course of the summer. This finding would clarify key findings by Cardoos et al. (2008) who emphasized youth/groups who engage in deviant talk, rather than youth/groups with externalizing problems. It would also extend beyond Kretsch et al.

Residential Iatrogenic Effects 31 (2008) who showed many person-group differences in peer group experiences for aggressive and non-aggressive kids in aggressive and non-aggressive groups, but few implications of these differences for treatment effects. Finally, gender is an important factor to examine in the study of iatrogenic effects in group treatment for externalizing youth. Most research on antisocial youth is based on boys, primarily because of prevalence differences in conduct disorder and delinquency, and differences in referral rates for treatment and intervention such that treatment programs are heavily populated by boys. For the most part, boys are more physically aggressive than girls (Keenan & Shaw, 1997). We know, however, when girls are externalizers, they are more often comorbid (with depression, anxiety, PTSD) than boys (Zahn-Waxler, 1993). So, if pure externalizing is the problem, it is possible that girls will not show as much negative influence or be as affected by it in the same way. Externalizing girls, however, have the potential to be particularly problematic based on research by Cardoos (2006) who showed evidence of higher support for antisocial girls at Wediko. Reinforcement of antisocial behavior could cause susceptibility to iatrogenic effects. Also, a peer group comprised of antisocial girls is rare. Most of the girls referred to Wediko are one of at most a couple girls in their special education classrooms at home. Wediko brings all of these girls together and provides peer support for otherwise outcast females. While it is good that the program encourages peer support, this novelty could also exacerbate negative peer influence that has been studied in boys. Summary. This research set out to examine six research questions. (1) Is residential treatment vulnerable to the iatrogenic effects described in the deviancy training literature? (2) How do externalizing youth in residential treatment groups affect

Residential Iatrogenic Effects 32 peer group dynamics? Specifically, do they engage in more antisocial influence and less prosocial influence? (3) How do externalizing youth differ from mixed internalizing/externalizing youth in their influence on peer dynamics? (4) How does the aggregation of externalizing youth influence treatment outcomes for youth in residential treatment? (5) Do externalizing, mixed, and non-aggressive youth respond differently when in high externalizing versus low externalizing groups? (6) Are there gender differences in peer group dynamics and does this influence treatment effects?

Residential Iatrogenic Effects 33

Method Participants Participants were 280 children attending Wediko’s summer program in Hillsboro, New Hampshire in the summers of 2006 and 2007. Sixty-nine percent were boys (n = 193), and 31% were girls (n = 87). Typically the ratio of boys to girls attending the summer program is around three to one, which is a reflection of the higher rate of overt behavioral problems and referral, particularly in prepubescent boys (Goodman, Lahey, Fielding, Dulcan, Narrow, & Reiger, 1997). The mean age for the sample was 13.06 years (SD = 2.80, range = 7-20 years old). Children at Wediko’s summer program come from all over the country, with a majority residing in the greater Boston area. The current sample of children is racially diverse, with a majority of Caucasian (56.1%), followed by 30% African American, 11.8% Hispanic, and 2% identified as “other.” Seventeen children (approximately 1%) were identified as biracial (11 African-American/Caucasian, 5 Hispanic/Caucasian, 1 Caucasian/Native American). All children were recommended to the program by teachers, parents, social service, agencies or other mental health professionals. Most children in the sample were funded through their school district or social services, while others were privately funded to attend Wediko. Funding status depends on referral source and individual conditions. Each child goes through a thorough application and interview process during which the child and family must express an interest in attending Wediko with the goal of change.

Residential Iatrogenic Effects 34 Children and youth referred to Wediko are behaviorally, socially, or emotionally at-risk and their problems interfere with their ability to function properly at home, school, or with peers. Problems include but are not limited to attention deficit disorders, mood disorders, impulse control problems, attachment disorders, oppositional defiant disorder, poor social skills, atypical social development, severe aggression, and post-traumatic disorder. According to parent reports on the Child Behavior Checklist (Achenbach, 1991) available for 84.6% of the sample (n = 237), aggression is the only overall clinical elevation for the sample, and attention problems, total problem behavior, and externalization are sub-clinically elevated (See Table 1). Setting The summer program is a 45-day residential program that includes both nonacademic and academic components. The therapeutic program is designed to promote self-control, development of social skills, age-appropriate behavior, and personal growth in a highly structured environment. As many children at Wediko find it difficult to adapt to change, the program follows the same schedule daily. Children engage in a routine of daily activities including wake-up routines, mealtimes, Think City (Wediko’s academic component), group therapy, cabin activities including instruction in water sports (e.g. swimming, fishing), field sports (e.g. archery, athletics), arts and crafts (dance, art, theater), and pre-vocational activities (e.g. kitchen help, landscaping). Children at Wediko are assigned to clinical groups with approximately 8 to 12 peers with whom they spend most of the day. The clinical groups are part of one of the four developmental programs which are based on chronological age, developmental level, and social strengths and weaknesses.

Residential Iatrogenic Effects 35 Each clinical team is also comprised of a variety of staff members, including on average three residential counselors who remain with the children at all times. Teams also include a Think City teacher, who spends most of the day teaching, but is often with the group for mealtimes and bedtime. Clinical teams also include activity staff who switch between being with the group and leading an activity throughout the day. The clinical team is headed by a supervisor who has an advanced degree, and generally many years of experience with Wediko Children’s Services. The supervisor does most of the planning for group activities, runs group therapy, and corresponds with parents of children within the group. Each team also has a lead staff member who has experience working in the summer program and is the day-to-day manager of the clinical team. Wediko staff live with the children in their clinical group, six days a week, for a large portion of each day. Think City teachers and activity staff spent at least 5 or more hours a day with the children, while residential counselors spent more than ten hours a day with the group of children. If the staff member is the same sex as the children, he or she also sleeps in the cabin at night, while staff of the opposite sex sleep in a dormitory style staff housing accommodation on setting. In addition to addressing the setting-wide goal to resolve conflict without aggression, the high child to staff ratio (approximately 2 to 1) allows for individualized intervention plans and high supervision at all times, and ample support from staff in encouraging personal socioemotional growth. Measures Teacher Report Form (TRF). The TRF (Achenbach and Rescorla, 2001, see Appendix A.) was completed three times throughout the summer for each child by

Residential Iatrogenic Effects 36 Table 1 Means and Standard Deviations for Narrow Band and Broad Band Scales on the Child Behavior Checklist by Parent Report

Mean

Standard

Deviation Anxiety

64.46

10.83

Withdrawal

64.11

10.08

Somatic

60.12

8.58

Social

67.77

9.11

Thought

65.16

8.60

Attention-Seeking

68.53

9.63

Rule Breaking

65.45

7.88

Aggression

70.51

10.34

Internalizing

64.21

9.55

Externalizing

68.21

8.52

Total Problem Behavior

68.05

7.83

n = 237

Residential Iatrogenic Effects 37 Wediko staff at two week intervals beginning on the tenth day of the program. The standardized scale contains 118 items to rate overall problem behavior levels of children in the program. Items include eight syndromes, or narrow band scales, (anxiety, aggression, withdrawal, delinquency, attention problems, somatic complaints, thought problems, and social problems), which are grouped into two broader categories (broadband scales) of internalizing (comprised of anxiety, withdrawal, and somatic behaviors) and externalizing behaviors (comprised of aggression and rule breaking). There is also an overall scale of total problem behavior including all narrow band scales. This widely used measure is reliable with a test-retest reliability of .82-.87 and short-term stability at .57-.83 (Achenbach & Rescorla, 2001). Average scores for TRFs completed by Wediko staff over the three time periods showed no clinical elevation in any narrow or broadband, but there was a subclinical elevation for aggression (See Table 2.). To create child behavior groups (child type) for analysis I used mean internalization and externalization for the TRF over the three time periods: I calculated mean levels of externalizing behaviors for both males and females using age and gender adjusted Tscores. The official clinical cutoff is 70, but the common borderline clinical cutoff is 65. Using a score of 65 as the threshold, participants were placed into one of four categories: (1) Children with a score of <65 on both internalization and externalization were considered non-clinical. (2) Children with a score of <65 on internalization and ≥65 on externalization were classified as pure externalizers. (3) Children with a score of ≥65 on internalization and <65 on externalization were considered internalizers. (4) Children with a score of ≥65 on both internalization and externalization were considered mixed.

Residential Iatrogenic Effects 38 The largest group of the sample was pure externalizers, followed by non-clinical children, mixed children, and internalizers (See Table 3). Peer assessments. At weeks two and five during the treatment program, peer assessments were conducted through sociometric interviews. These interviews were conducted individually by staff on the child’s clinical team. Each interview lasted approximately 25-30 minutes. The peer assessment interview was developed in previous research at Wediko (Wright et al., 1986, see Appendix B.) and includes assessments of social acceptance and rejection, prosocial behaviors, antisocial behaviors, social withdrawal, social experiences (both prosocial and antisocial), and peer influence (prosocial and antisocial). Children made nominations for questions like “Who do you really like?,” “Who stays by himself/herself a lot?,” “Who talks in a friendly way to other kids a lot?” To answer these questions, each child received a stack of photos of children in their clinical group. Each photo had an identification number on the back to maintain confidentiality of the child’s nominations. Children were discouraged from talking about their nominations with peers, so that each child’s responses were unique and uninfluenced. For each item, the child placed photos of their peers that they felt fit the description face down, and the interviewer recorded the numbers on the back of the photos. Nominations were unlimited. Peer assessments were scored during the summer to determine perceived versus actual status within the group, as well as perceived versus actual rejection of peers within the group. Actual social status was also compared to levels of aggression and prosocial behavior within the group. Staff members used these results from the peer assessments to evaluate group dynamics and behavioral trends within clinical groups.

Residential Iatrogenic Effects 39 Table 2 Means and Standard Deviations for Narrow Band and Broad Band Scales on the Teacher Report Form by Wediko Staff Report Averaged Over Three Assessments

Mean

Standard

Deviation Anxiety

62.47

7.40

Withdrawal

61.57

6.91

Somatic

56.24

6.02

Social

64.61

6.56

Thought

60.56

6.19

Attention-Seeking

60.37

6.58

Rule Breaking

63.70

6.47

Aggression

68.22

7.33

Internalizing

64.97

5.59

Externalizing

66.96

6.30

Total Problem Behavior

62.50

6.68

n = 280

Residential Iatrogenic Effects 40 Table 3 Child Type Behavior Group as Determined by Internalizing and Externalizing Behaviors on TRF

Child TypeBehavior Group Int (SD) Mean Ext (SD)

n

Percent

Mean

Formatted: Font: Italic Formatted: Font: Italic Formatted: Font: Italic

Non-clinical

72

25.7

58.55 (4.78)

61.17 (3.10)

Boys

64

33.2

58.58 (4.79)

61.01 (3.23)

Girls

8

9.2

58.30 (5.02)

62.22 (1.53)

104

37.1

58.59 (5.00)

70.822 (4.25)

Boys

59

30.6

58.60 (4.84)

69.41 (2.99)

Girls

45

51.7

58.58 (5.26)

72.67 (4.92)

Internalizers

42

15.0

69.66 (2.98)

60.97 (3.06)

Boys

34

17.6

69.35 (2.60)

60.90 (3.08)

Girls

8

9.2

70.99 (4.22)

61.27 (3.20)

62

22.1

68.78 (3.10)

71.27 (5.00)

Boys

36

18.7

68.69 (2.69)

70.81 (4.60)

Girls

26

29.9

68.92 (3.64)

71.91 (5.52)

Externalizers

Mixed

Residential Iatrogenic Effects 41 Seven behavior scales were created from the Peer Assessment using items averaged over the two time points. Antisocial influence was assessed by two items (“Who likes to talk about breaking the rules?” and “Who thinks it’s funny or cool when kids break the rules or get into trouble?”) based on prior research by Dishion et al. (1996) regarding deviancy training and rule-breaking conversation in male youth dyads. Prosocial influenced was assessed by one item (“Who tries to help other kids follow the rules and stay out of trouble?”). Antisocial behavior was assessed by four items (“Who teases, bosses, or threatens other kids a lot?”, “Who hits or pushes other kids a lot?”, “Who tries to stop other kids from joining in and being part of the group?”, and “Who says mean things about other kids when they’re not around?”). Prosocial behavior was assessed by three items (“Who talks in a friendly way to other kids a lot?”, “Who gets along with others and cooperates a lot?”, and “Who follows the rules and tries not to get into trouble?”). Withdrawal was assessed by three items (“Who stays by him/herself a lot?”, “Who looks sad or cries a lot?”, and “Who whines or complains a lot?”). Antisocial experience was assessed by four items (“Who gets teased, bossed, or threatened a lot by other kids?”, “Who gets hit or pushed a lot by other kids?”, “Who do other kids try to stop from joining in and being part of the group?”, and “Who do other kids say mean things about when he/she is not around?”). Prosocial experience was assessed by two items (“Who do other kids get along with and cooperate with a lot?” and “Who do other kids talk to a lot in a friendly way?”). See Table 4 for item intercorrelations and alphas for each scale. Clinical Global Impressions (CGI). The CGI used at Wediko is modeled loosely after the Clinical Global Impressions scale developed by Guy (1976), which is used to

Residential Iatrogenic Effects 42 assess treatment response on an index of global severity and symptom-specific improvement among psychiatric patients. Wediko’s CGI is similar in that it uses a sevenpoint Likert scale ranging from 1 (very much worse) to 4 (no change) to 7 (very much improved) and it assesses overall change in behavior over the course of the treatment program as is done on the central item of the original CGI (See Appendix C). Similar adaptations have been used in other studies of residential treatment response (Connor et al., 2002). For the Wediko assessment all staff assigned to a clinical group assessed each child in the group and average improvement scores were obtained to minimize the effects of individual rater bias. Additional items included questions to assess changes in frequency of aggressive, withdrawn, and prosocial behaviors. There were also items to assess changes in the child’s interactions with peers and adults. Staff members completed the CGI at the last data collection period, which occurred during staff reorientation after the children had completed treatment. CGI assessments completed by staff showed an overall trend of positive behavior change on all scales (global change, adult interaction, peer interaction, aggression, withdrawal, and prosocial behaviors). (See Table 5; two participants had missing data on the CGI.). Procedures Data were collected over two summers from Wediko summer staff, parents, and teachers. Wediko counselors completed the TRF assessments three times over the course of the summer (at weeks two, four, and six). At each assessment, staff rated behavior over the last two weeks. Each child was rated by two different staff members (i.e. a residential counselor, activities

Residential Iatrogenic Effects 43 counselor, or Wediko teacher). Staff members were randomly assigned to the children they rated at Time 1. They rated the same children at Time 3. At Time 2 they were randomly assigned to rate a different child in their group. They rated the same child at Time 2 and Time 4 (when the only assessment was the CGI). For these analyses, the rater who had spent the most time with the child over the two week period was selected at each time point as the primary rater. Only primary ratings were used in these analyses. Data collection and management were coordinated by the Wediko summer research team, comprised of four to five clinical research staff. Members of the research team were also part of a clinical team, but spent several hours each day away from the developmental group preparing, processing, and organizing data and measures. They were supervised by research consultants and directors of the Summer Program. All parents/guardians of children attending Wediko were aware of the clinical assessment program prior to sending their child to the New Hampshire summer program. They were informed that research would be conducted throughout the summer to help tailor programming and interventions to best suit their child and the group. Furthermore, parents/guardians were informed that all data collected would remain confidential, and that data would never be used to single out an individual child. It was stated that data collected during the summer was for clinical purposes only. All participants in the present study had parental consent to participate in research (see Appendix D).

Residential Iatrogenic Effects 44 Table 4 Peer Assessment Subscale Reliability and Intercorrelations

Scale name

# of items

α

r range

Antisocial influence

2

.86

.92

Antisocial behavior

4

.51 - .78

.90

Prosocial behavior

3

.77 - .85

.92

Withdrawn behavior

3

.39-.63

.77

Antisocial

4

.51 - .77

.90

2

.83

.90

experience Prosocial experience

Note: Prosocial influence has only 1 item, so it is not included in this table. For scales that contained two items, r range represents the pairwise item correlations. For scales that contained more than two items, multiple r’s were calculated.

Residential Iatrogenic Effects 45 Table 5 CGI Means and Standard Deviations by Aggregated Staff Report

Improvement Scale

Minimum

Maximum

Mean

Std.

Global

3.00

7.00

5.53

.75

Adult Interact

2.88

6.75

5.43

.64

Peer Interact

3.11

6.88

5.36

.66

Aggression

2.38

7.00

4.67

.78

Withdrawal

2.78

6.88

4.74

.71

Prosocial

3.00

7.00

5.30

.64

Dev.

Residential Iatrogenic Effects 46 Results Differences in Antisocial and Prosocial Behaviors by Child Type The following analyses address whether residential treatment is vulnerable to iatrogenic effects as the deviancy training literature describes. These analyses also examine how externalizing youth contribute to antisocial norms and detract from prosocial norms. To assess whether children with different behavior profiles differentially promote antisocial norms, a 2 (gender) X 4 (child type) analysis of variance conducted across four behavior groups (non-clinical, pure externalizers, internalizers, and mixed internalizers/externalizers) was conducted averaging scores on the antisocial influence subscale over both available peer assessment points. It revealed significant differences in overall antisocial influences by behavior group, F(3, 271) = 15.379, p < .05, η2 = .195. Post Hoc Tukey tests revealed that pure externalizers showed the highest rate, and were significantly higher for antisocial influence than non-clinical, internalizers, and mixed children (See Figure 1). Follow-up analyses were conducted separately on the two variables that define the antisocial influence scale (talking about breaking the rules, laughing when others break the rules), which yielded similar findings. There were no significant main effects or interactions involving gender in any of these analyses. Age was not included as a factor in these analyses due to small cell sizes with all factors included. Preliminary analyses suggested no age interactions with child type for antisocial influence. To assess whether children with different behavior profiles differentially undermine

Residential Iatrogenic Effects 47 Figure 1. Means of antisocial influence for non-clinical, externalizers, internalizers, and mixed behavior children

Proportion of Nominations

0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 Non-Clin

Ext

Int

Mixed

Residential Iatrogenic Effects 48 prosocial norms, a 2 (gender) X 4 (child type) analysis of variance was conducted across four behavior groups (non-clinical, pure externalizers, internalizers, and mixed internalizers/externalizes). It revealed a significant main effect for behavior group, F(3, 271) = 3.466, p < .05, η2 = .114. There were no effects involving gender. Post Hoc Tukey tests revealed that nonclinical children (M = .354) showed significantly higher rates of prosocial influence than pure externalizers (M = .285) and mixed internalizers/externalizers (M = .277). Furthermore, internalizers (M = .407) also showed significantly higher rates of prosocial influence than both externalizing groups. Both pure externalizers and mixed children were significantly lower contributors to prosocial norms, but did not differ from one another. To further examine the promotion of antisocial norms, a parallel ANOVA investigating aggregated aggression by peer report (physical, verbal, and social) was conducted. It yielded significant results for behavior group, F(3, 271) = 13.054, p < .05, η2 = .210. Both externalizing groups showed significantly higher rates of aggression (pure externalizers M = .317, mixed M = .293) than both the non-clinical (M = .194) and internalizing children (M = .119). There were no significant main effects or interactions with gender. To further examine possible undermining of prosocial norms, a 2 (gender) X 4 (child type) analysis of variance was conducted on overall prosocial behavior throughout the summer (talks in a friendly way, gets along and cooperates, follows the rules). There was a significant main effect for behavior group, F (3, 271) = 13.242, p < .05, η2 = .228. Post Hoc Tukey tests revealed internalizers exhibited significantly more prosocial behaviors

Residential Iatrogenic Effects 49 (M = .582) than both externalizing groups (pure externalizers M = .327, mixed M = .343). Similarly, non-clinical children also displayed significantly more prosocial behaviors (M = .422) than both externalizing groups. Mixed internalizers/externalizers did not differ significantly from externalizers. There was also a significant interaction effect for gender by behavior group F (3, 271) = 3.080, p < .05, η2 = .228. Simple effects tests revealed that non-clinical boys were higher than both pure externalizers and mixed internalizers/externalizers, but this was not true for girls. All three girls groups showed similarly low levels of prosocial behavior by peer report. Also of note, internalizing girls were especially high in prosocial behavior. Because this variable is the main process variable identified in the previous studies of peer contagion and iatrogenic treatment effects for antisocial youth, I focus in the next analyses on the aggregation of externalizing youth – the ones who are most likely to actively promote antisocial norms. Summary Overall, externalizers and mixed internalizers/externalizers showed similar antisocial and prosocial behavior within their peer groups. Both engaged in less prosocial behavior and more antisocial behaviors. Both also engaged in less encouragement of prosocial norms. However, externalizers were most likely to actively encourage antisocial behavior.

Differences in Antisocial and Prosocial Influence and Behavior by Child and Externalizing Group Type Preliminary Analyses. To investigate whether aggregation of externalizing children (who engaged in the most antisocial influence) affects peer group behavior and

Residential Iatrogenic Effects 50 peer group experiences for different types of children in treatment, I sought to identify clinical treatment groups that were low or high in the proportion of externalizing group members. I first examined the proportions of externalizing children in each clinical group. Using the same behavior grouping standard that was used for each child individually, each clinical group was identified as either “high externalizing” or “low externalizing” depending on the ratio of pure externalizers to other children within the group. At first, groups with 33% or more externalizers were classified as “high externalizing.” Previous research (Dodge et al., 2006) suggests that one or two antisocial youth in a group may not be enough to create problematic peer dynamics, but beyond this point the risk increases sharply. The 33% membership rule, however, resulted in more girls’ groups being classified this way than boys’ groups. To avoid a gender confound, it was necessary to use different standards for the formation of externalizing groups for boys and girls. Using the greater than or equal to 33% cutoff, all but one of the nine girls’ groups were identified as high externalizing. When the cutoff standard was raised to greater than or equal to 50%, five were identified as high externalizing and 4 were not (See Table 6). For boys, the cutoff was kept at 33% yielding 8 high externalizing groups and 11 low externalizing (See Table 7). A 50% cutoff would have created too few high externalizing groups (only 2). As a result of the disparity in the proportion of externalizing children, boys and girls were analyzed separately in subsequent analyses. Externalizing and non-externalizing groups differed in age for both genders. Both non-externalizing boys and girls groups were older. This reflects both referral patterns and Wediko’s admission practices. Typically, Wediko does not admit extremely

Residential Iatrogenic Effects 51 aggressive older boys and girls for safety reasons. Also, highly aggressive older boys and girls are more often referred to juvenile detention centers than to mental health services. All subsequent analyses control for age by using age as a covariate. Finally, because there were very few non-clinical (n = 9) and internalizing ( n = 10) girls in the sample, these groups were collapsed to create one “non-externalizing group.” This was done over both genders to create parallel analyses. Early analyses revealed no significant differences between non-clinical and internalizing groups on any subscales. All subsequent analyses regarding child type will refer to one of three groups: non-externalizing, externalizing, and mixed. Boys. To examine whether group composition affects antisocial and prosocial influences, as well as other peer behavior for different types of children in treatment, a 3 (child type) by 2 (group type) by 2 (time) repeated measures ANOVA on both available assessments was conducted on each with age as a covariate. For antisocial influence, there was a significant group effect over time, Wilks’ lambda = .859, F(1,181) = 29.621, p < .001. Boys in high externalizing groups showed significantly increased levels of deviant talk from early (M = .200) to late summer (M = .364). For low externalizing boys’ groups, there was a slight but significant decrease in deviant talk from early (M = .291) to late (M = .274). There was also a significant three way interaction for child type by group type over time, F(1,181) = 3.99, p < .02 All child types showed a significant increase in deviant talk when they were in externalizing groups, but mixed kids showed an especially sharp increase in deviant talk from early to late summer if they were in externalizing groups (See Figure 2). Mixed boys also showed a decrease (marginally significant in pairwise comparisons) if they were in low

Residential Iatrogenic Effects 52 externalizing groups. In contrast, non-aggressive and externalizing boys increased their deviant talk, although not significantly, even if they were in non-externalizing groups. Between-subjects effects showed identical child type effects to preliminary analyses, F(2, 181) = 18.248, p < .001, where externalizing boys showed significantly elevated levels of deviant talk overall (M = .391) compared to non-externalizing boys (M = .195) and mixed boys (M = .260). For prosocial influence there was only a significant between subjects child type effect, F(2, 181) = 14.108, p < .001, which mirrored the results found in preliminary analyses. Girls. To parallel the analyses for boys, a 3 (child type) by 2 (group type) by 2 (time) repeated measures ANOVA was conducted on antisocial and prosocial influence with age as a covariate. For antisocial influence, there was a significant effect for group type over time, that mirrored the effect found for boys, Wilks’ lambda = .872, F(1, 77) = 11.306, p < .001. Girls in high-externalizing groups increased in antisocial influence from early (M = .199) to late (M = .302) summer, while girls in low-externalizing groups showed decreased antisocial influence from early (M = .285) to late (M = .235). There was also an expected between subjects child type effect that mirrored the preliminary analyses done with all children, F(2, 77) = 8.135, p < .001, where externalizing girls showed significantly higher rates of antisocial influence (M = .377) compared to non-externalizing girls (M = .150) and mixed girls (M = .238). Time was the only significant factor for prosocial influence, Wilks’ lambda = .855, F(1, 77) = 13.029, p < .001. Girls showed significantly increased prosocial influence later

Residential Iatrogenic Effects 53 Table 6 Girls’ Clinical Groups Externalizing Proportions

High Ext

Group Name

Number of Externalizing

Proportion of

Externalizing 06 Chipmunks

7

78

06 Esperanza

5

50

06 Pegasus

8

72

07 Bluefish

6

75

07 Walkers

5

50

06 Walkers

4

36

07 Artemis

4

40

07 Esperanza

3

30

07 Pegasus

4

40

Low Ext

Residential Iatrogenic Effects 54 Table 7 Boys’ Clinical Groups Externalizing Proportions

High Ext

Group Name

# of Externalizing

Proportion of Externalizing

06 Falcons

4

40

06 Longhorns

5

45

06 Woodchucks

3

33

07 Orion

4

44

07 Saturn

6

60

07 Woodchucks

4

40

07 Arthur Ashe

4

40

07 Falcons

4

40

06 Apollo

3

25

06 Daedalus

1

9

06 Neptune

2

18

06 Orion

2

20

06 Otters

3

27

06 Saturn

3

30

07 Daedalus

2

17

07 Neptune

1

9

07 Otters

3

30

07 Apollo

1

10

07 Monarchs

3

30

Low Ext

Residential Iatrogenic Effects 55

Peer Reported Antisocial Influence

Figure 2. Boys’ change in antisocial influence by child and group type.

0.5 0.4

NonCh-LoExGp NonCh-HiExGp ExCh-LoExGp ExCh-HiExGp MxCh-LoExGp MxCh-HiExGp

0.3 0.2 0.1 0 1

2 Time

Residential Iatrogenic Effects 56 in the summer (M = .390) compared to earlier (M = .297). Summary. Overall, group composition had a significant impact on behavior. Both boys and girls in high externalizing groups consistently showed more dramatic increases in antisocial influence compared to children in low-externalizing groups. For boys’ groups, however, the significant three way interaction suggests that child type is also a significant factor. Mixed boys were especially likely to be influenced by group compositions, showing sharp increases in antisocial influence when in externalizing groups, and decrease in antisocial influence when in non-externalizing groups. Externalizing and non-externalizing boys both showed increases in antisocial influence when in high-externalizing groups as well. These boys also showed little change when in low-externalizing groups. Although this interaction was not observed for girls, it is important to take into consideration the extremely high proportion of externalizing girls in the sample. “Low externalizing” girls’ groups had many externalizing girls, sometimes up to just under 50%. Although I still detected a group type effect, the child by group interaction may have been influenced by the weaker differentiation between low and high externalizing groups. Also, power to detect a three way interaction was lower in the smaller sample of girls. Finally, group composition had little impact on either boys or girls’ prosocial influence, which suggests that person-group interactions are noticeably influential for antisocial behaviors.

Overall Behavior Change Using TRF To assess whether the aggregation of externalizing youth affected outcome, as well as group process (antisocial influence), a series of repeated measures ANOVAs on

Residential Iatrogenic Effects 57 all three available time points were conducted using relevant broad- (internalizing and externalizing) and narrow-band scales of the TRF with child type X group type and age as a covariate. Because different cut-offs were used to define group type for boys and girls, all of the following analyses were conducted separately for the two genders. Boys First, a repeated measures ANOVA was conducted with total problem behavior as a repeated measure for child type X group type with age as a covariate. There was a significant interaction for child group X child type over time, Wilks’ lambda = .942, p < .05. As shown in Figure 2, non-externalizing children in high-externalizing groups showed the most change in problem behavior over the three evaluation points. Follow-up testing revealed that they were the only group that showed a significant increase in multivariate testing. Pairwise testing revealed that total problem behavior differed between T1 and T3 for these kids. The only other significant pairwise comparison was for externalizing youth in non-externalizing groups who decreased in total problem behavior from T2 to T3. Pairwise tests of each child type in externalizing versus nonexternalizing groups at each of the three time points yielded only one difference, for the non-externalizing children at T3 (See Figure 3). There was also a significant effect for time, Wilks’ lambda = .947, p < .01. Boys showed increased total behavior problems over time, (M1 = 64.842; M2 = 65.751; M3 = 65.488), but the interaction discussed previously qualifies this main effect. There was also a significant interaction for age over time, Wilks’ lambda = .945, p < .01, where younger boys showed elevated problem behavior at T2 compared to T1 and T3, which were similar. Older boys showed increased problem behavior over time.

Residential Iatrogenic Effects 58 There was an expected significant between subjects main effect for child type, F(2, 180) = 78.592, p < .001, where mixed internalizing/externalizing children showed the most total problem behavior. There was also a significant between subjects main effect for child group, where externalizing groups showed significantly more problem behavior (M = 66.191) compared to non-externalizing groups (M = 64.53). To localize the behavior change identified in the total problem behavior analysis, parallel repeated measures ANOVAs were planned for the two broad band measures of Externalizing and Internalizing, with appropriate follow-up tests using narrow band component scales where appropriate. For externalizing behaviors, there were significant changes in externalizing behaviors over time for child age, Wilks’ lambda = .946, p < .001, and child type, Wilks’ lambda = .948, p < .05, and for the interaction between child type and child group over time, Wilks’ lambda = .933, p < .05 (See Figure 4). Two groups demonstrated significant multivariate change, non-externalizing boys in externalizing groups who showed a steady increase, and externalizing boys in nonexternalizing groups who significantly increased from T1 to T2, and then significantly decreased T2 to T3. A pattern very similar to this latter one was observed for nonexternalizing boys in non-externalizing groups with significant T1 to T2 and T2 to T3 comparisons, but no significant multivariate contrast. Comparing the different child types in externalizing versus non-externalizing groups, non-externalizing children differed at T3, and so did externalizing children. Externalizing and mixed children differed over group types at T1 (See Figure 3). This interaction qualified the significant (nonlinear) effect for time, Wilks’ lambda = .940, p < .01, where externalizing behaviors significantly increased from T1 (M = 66.533) to T2 (M = 67.820), and remained high at

Residential Iatrogenic Effects 59 T3 (M = 67.224). There was also an interaction between time and the covariate of age, and the expected main effects for both child type and child group that paralleled those found in the Total Problem Behavior Analyses. To better understand these changes in externalizing behaviors, repeated measures ANOVAs were run on rule-breaking and aggression, the two narrow- band scales that comprise the externalizing scale. Unlike for total behavior problems and the externalizing broad band scale, there were no interactions between time and child type, group type, or their interaction for the narrow band component of rule-breaking. There were, however, significant (nonlinear) changes over time overall, Wilks’ lambda = .934, p < .01, where rule-breaking behaviors increased from T1 (M = 62.760) to T2 (M = 63.90), and then decreased again at T3 (M = 62.909), but to a score that did not differ from T1 or T2. The covariate age was also significant where younger boys showed more rule-breaking behavior (M = 63.473) compared to older boys (M = 60.995). There was a predicted significant child type effect, where externalizing boys showed the highest levels of rulebreaking behavior (M = 65.877) followed closely by mixed boys (M = 65.200) and nonexternalizing boys showed the lowest levels of rule-breaking (M = 58.492). For aggression, the second component of the externalizing broad-band scale, child type by group type over time was significant, Wilks’ lambda = .913, F(4, 358) = 4.166, p < .01. The overall pattern was similar to the pattern described for Externalizing. Two groups demonstrated significant multivariate change, non-externalizing boys in externalizing groups who showed a steady increase, and externalizing boys in nonexternalizing groups who significantly increased T1-T2 and then significantly decreased T2-T3. The pattern for non-externalizing boys in non-externalizing groups was weaker

Residential Iatrogenic Effects 60 Figure 3. Boys’ change in total problem behavior by child and group type.

TRF TPB T-Score

80 75 NonCh-LoExGp NonCh-HiExGp ExCh-LoExGp ExCh-HiExGp MxCh-LoExGp MxCh-HiExGp

70 65 60 55 50 1

2 Time

3

Residential Iatrogenic Effects 61

TRF Externalizing T-Score

Figure 4. Boys’ change in externalizing by child and group type.

80 75 NonCh-LoExGp NonCh-HiExGp ExCh-LoExGp ExCh-HiExGp MxCh-LoExGp MxCh-HiExGp

70 65 60 55 50 1

2 Time

3

Residential Iatrogenic Effects 62

TRF Aggression T-Score

Figure 5. Boys’ change in aggression by child and group type.

80 75 NonCh-LoExGp NonCh-HiExGp ExCh-LoExGp ExCh-HiExGp MxCh-LoExGp MxCh-HiExGp

70 65 60 55 50 1

2 Time

3

Residential Iatrogenic Effects 63 with a significant T1-T2 increase only. Comparing the different child types in externalizing versus non-externalizing groups, non-externalizing children differed at T3. Externalizing and mixed children differed over group types at T1. This interaction qualified the significant (nonlinear) main effect for time, Wilks’ lambda = .957, p < .05, in which boys increased their aggression from T1 (M = 68.197), to T2 (M = 69.684), and remained high at T3, but not different from T1 or T2. See Figure 5 for a summary of means. Age was also a significant factor where younger boys showed significantly more aggressive behaviors than older boys. For internalizing behaviors there were no significant time effects. There was only an expected significant between child type effect where mixed children showed the highest rates of internalizing behaviors followed by non-externalizing and externalizing boys. Summary. The aggregation of externalizing youth had its most obvious effect on non-externalizing children’s responses to treatment. For total behavior problems, externalizing, and aggression nonclinical boys in externalizing groups showed reliable increases in problem behaviors. For all of these behaviors the effect of being in a highly externalizing group was gradual, with nonclinical boys in externalizing groups differing from those in non-externalizing groups only at T3. Externalizing boys were also affected by group membership, but less dramatically. For total behavior problems, externalizing and aggression externalizing boys decreased their problem behavior from T2 to T3. Often this was after a brief period of escalation from T1 to T2. Overall, boys’ total problem behavior showed signs of increase over the summer, but this was not always linear, and it was often qualified by interactions with child type and group type described above.

Residential Iatrogenic Effects 64 Girls The same analysis plan used for boys was followed for the girls. A repeated measures ANOVA was conducted with total problem behavior as a repeated measure for child type X group type with age as a covariate. There was a significant effect for time X child type, Wilks’ lambda = .864, F(4, 148) = 2.795, p < .05, where most notably nonexternalizing girls showed increased problem behavior over time with a significant increase between T1 and T3. Externalizing girls showed change in problem behavior peaking at T2 (M = 69.52), but returning to a lower level at T3 (M = 67.081). T1 (M = 66.953) levels did not differ from T3 levels. Mixed girls did not show a reliable decline in problem behavior. There was an expected between subjects main effect for child type with mixed girls showing the highest rates of total problem behavior overall. There was also a significant interaction for child type X group type, F(2, 75) = 3.461, p < .05. Nonexternalizing girls in non-externalizing groups showed significantly less problem behavior overall than non-externalizing girls in externalizing groups. For mixed girls, there was more reported problem behavior in low-externalizing groups than in highexternalizing groups. Externalizing girls were not reliably affected by group membership. For externalizing behaviors, time was not a significant factor. The only significant factor was child type, F(2, 75) = 26.532, p < .001, which was expected. Externalizing girls (M = 72.663) and mixed girls (M = 71.525) had higher overall levels of externalizing behaviors than non-externalizing girls (M = 61. 939). To parallel analyses for the boys, repeated measures ANOVAs were run on the two narrow- band scales that comprise the externalizing scale. For rule-breaking, time was a significant factor for child type, Wilks’ lambda = .858, F(4, 148) = 2.938, p < .05,

Residential Iatrogenic Effects 65 where non-externalizing girls showed increased rule-breaking over time. Most notably externalizing girls showed a peak at T2 then a significant drop to T3 that resembled the pattern for total behavior problems. Mixed and non-externalizing girls did not show reliable change. A predicted main effect for child type revealed that externalizing and mixed girls showed significantly higher rates of rule-breaking than non-externalizing girls. For aggression, time was not a significant factor. The only significant main effect was expected, for child type. This effect was identical to the effect for rule-breaking and externalizing. For internalizing behaviors, there was a significant effect for child type over time, F(4, 148) = 3.51, p < .01, where most notably mixed girls showed a significant decline in internalizing symptoms with a significant decrease from T1 to T3. There was also an anticipated child type effect, F(2, 75) = 25.39, p < .001, with mixed girls (M = 68.91) demonstrating more internalizing symptoms overall than the non-externalizing girls (M = 64.94) who showed more than the externalizing girls (M = 59.51). Summary. For girls, there was less power to detect 3-way interactions and less evidence of group type influence. There was, however, a significant increase in total problem behavior for non-externalizing children, and a late summer (T2-3) drop in total problem behavior for externalizing girls. Both of these findings paralleled key findings for boys, but without the interacting factor of group type. Group type affected overall levels of behavior problems, but not change for girls. Non-externalizing girls in low externalizing groups showed fewer behavior problems than their counterparts in high externalizing groups, whereas the reverse was true for mixed girls. Externalizing girls were less affected by group membership.

Residential Iatrogenic Effects 66 Overall Change. In a final set of analyses, global change measures were analyzed to assess how child type and group type affected overall improvement. The CGI used at Wediko is a measure of change, but it is based on impressions of change rather than actual measures of change over several time periods. With the data from Wediko, perceived change according to the CGI is assessed on one point in time, but counselors are asked to think back over the entire summer. Scores are derived from averaging all counselors in the group to remove individual rater bias, making it a reliable overall assessment of impressions of improvement. Such measures may tap into more than just changes in behavior problems, and may be affected by changes in social relationships, prosocial behavior, and adaptive functioning. Although the TRF may be more objective, the CGI offers a different view of treatment response that can complement traditional analyses of behavior change. Analyses are focused on the broadest dimensions of change assessed in the CGI – and those that do not directly overlap with the TRF analyses. To examine whether the aggregation of externalizing children affects broad treatment outcomes, a 3x2 MANOVA was conducted on the CGI (Clinical Global Impressions) global change, change in relationships with adults, and change in relationships with peers variable for both boys and girls separately with age as a covariate. Boys. A MANOVA was conducted across the three child types (non-aggressive, externalizers, and mixed) and two group type levels (high externalizing and lowexternalizing) that revealed a multivariate effect for child type, Wilks’ lambda = .931, F(6, 368) = 2.23, p < .05, and a significant child type by group interaction, Wilks’ lambda = .932, F(6, 368) = 2.186, p < .05. Univariate tests revealed no significant child type main effects for global, adults or peers alone. Univariate tests did reveal, however,

Residential Iatrogenic Effects 67 that there was a significant interaction between child type and group type for the global scale, F(2, 186) = 3.506, p < .05, where non-externalizing boys in low-externalizing groups showed more global change than non-externalizing boys in high-externalizing groups. In contrast, for mixed internalizing/externalizing boys, those in highexternalizing groups showed significantly higher rates of global change than mixed boys in low-externalizing groups. In contrast to peer contagion hypotheses, externalizing boys were affected very little by group membership with boys in low-externalizing groups showing slightly less global change than externalizing boys in externalizing groups (See Figure 6). For changes in relationships with adults, there was also a significant interaction between child type and group type, F(2, 186) = 3.207, p < .05. Non-externalizing boys in low-externalizing groups showed significantly more improvement in adult interaction (M = 5.467) than non-aggressive boys in high-externalizing groups (M = 5.127). For externalizing boys, there was no difference over group type. For mixed internalizing/externalizing boys, those in high-externalizing groups unexpectedly showed slightly higher rates of change (M = 5.659) than mixed boys in low-externalizing groups (M = 5.365). There was a similar pattern for global change. There were no significant changes in peer interactions for boys. Girls. A MANOVA was conducted across the three child types (non-aggressive, externalizers, and mixed) and two group type levels (high-externalizing and lowexternalizing) on the same set of CGI variables (global, adults, peers). There was a significant main effect for child type, Wilks’ lambda = .85, F(6, 152) = 2.16, p < .05. There was also a significant main effect for the covariate age. Univariate tests revealed

Residential Iatrogenic Effects 68 only a significant child type effect for peer relationships, F(2, 78) = 3.754, p < .05, where externalizing girls showed less change (M = 5.18) than non-aggressive girls (M = 5.61) and mixed internalizing/externalizing girls (M = 5.60). Summary Using global impressions of overall improvement and improvement in relationships with adults and peers, aggregating externalizers does affect treatment outcomes to some extent, but not necessarily in the way that previous research has suggested. As with TRF analyses, results were stronger for boys. Non-externalizing boys again seem most affected, and most negatively so, by exposure to externalizing peer groups. They showed significantly less improvement overall and in relationships with adults when in externalizing groups. Their mean improvement, although dampened, was still in the “improved” range. Interestingly, mixed children showed the unexpected pattern of improving more when in groups with externalizing youth. This was most notable in their global improvement. Mixed youth in non-externalizing groups were still in the “improved range,” but those in externalizing groups were higher. Externalizing boys were relatively unaffected in their overall improvement by group membership. For girls, there was only one notable effect. Externalizing girls improved less in their peer relations than did nonclinical and mixed girls. This effect did not depend on group membership.

Residential Iatrogenic Effects 69 Figure 6. Global change as measured by the CGI for child type X behavior group for boys.

6

CGI - Global Change

5.8

5.6 Low-Ext Hi-Ext 5.4

5.2

5 Non

Ext

Mixed

Residential Iatrogenic Effects 70 Discussion Controversy remains over whether or not residential treatment is a breeding ground for deviant peer influences. On the one hand, short-term, highly-supervised residential treatment has shown positive outcomes for children, particularly when positive gains are made in the home environment as well (McCurdy and McIntyre, 2004). Furthermore, there is no body of research that suggests residential treatment is particularly harmful. On the other hand, there is enough evidence from other treatment settings to cause concern, and there has been speculation that the aggregation of deviant peers in residential treatment is not conducive to positive treatment effects (Dodge, Dishion, and Lansford, 2006). It is therefore important to examine youth in residential treatment to determine whether children are harmed by deviant peer aggregation, which children are most harmful to group dynamics and which are most vulnerable to contagion effects in peer groups. The present study aimed to examine the ways externalizing children are problematic for non-externalizing peers in treatment with them. Participants were youth ages 7 to 18 with mixed behavior problems, including mostly serious behavioral and emotional disturbances. No research has been done examining differences between pure externalizers and mixed internalizers/externalizers, particularly in a residential treatment setting like Wediko. Mixed youth are often overlooked or associated with pure internalizing or pure externalizing groups. This distinction between externalizing and mixed internalizing/externalizing was therefore a focus of the present study. Children were classified as one of three behavior types based on staff-reported TRF measures. Pure externalizers were high on both rule-breaking and aggressive behaviors. Mixed

Residential Iatrogenic Effects 71 internalizers/externalizers were high on rule-breaking, aggression, withdrawal, anxiety, and depression. Non-externalizing children were those who did not meet the clinical cutoff on either the externalizing or internalizing broad-band scales on the TRF, or children who showed elevated levels of internalizing behavior. The latter two groups were collapsed into one after preliminary analyses because the focus of the present study was externalizing youth, particularly the differences between pure externalizers and mixed internalizers/externalizers. The numbers of pure internalizers, especially among girls, were also quite small. Because externalizing youth were expected to be more potentially harmful to peer dynamics (see below for findings that confirmed this hypothesis), clinical groups were defined as either high externalizing or low externalizing to gauge children’s exposure to deviant peer influence. This was done by calculating the proportion of pure externalizers in each group. Once the child type and group type were defined, I was able to examine both child type and group type effects on peer behavior and treatment response. For peer behavior, antisocial behavior and influence as well as prosocial behavior and influence were compared across two time points. For treatment response, behavior problems ratings over three time points were examined on the TRF and global impressions of improvement by staff as reported on the CGI at the end of the summer program were examined. As noted, special attention was paid to differences between the pure externalizers and mixed children. Child Behavior Type and Group Composition Effects on Peer Behavior As predicted by their group membership, externalizing youth (both pure and mixed externalizers) engaged in significantly more deviant behavior and less prosocial behavior compared to non-externalizing youth. Initially, it seemed that both pure and mixed

Residential Iatrogenic Effects 72 externalizers were cause for concern. Both demonstrated more antisocial behavior and less prosocial behavior within their groups. Neither group encouraged prosocial norms. However, I found that pure externalizers were the ones doing most of the antisocial influence in the form of deviant talking. This was true for both components of this behavior: talking frequently about breaking the rules and getting into trouble, and laughing and smiling when other kids broke the rules or got into trouble. Mixed children engaged in more deviant talk than non-externalizing children, but pure externalizers showed significantly higher rates of antisocial influence and antisocial behavior, and in doing so were the most likely to actively promote iatrogenic effects. For boys, being in a high externalizing group affected child engagement in deviant talk. Boys in high externalizing groups increased deviant talk overall, while boys in low-externalizing groups decreased somewhat. This group type effect was also seen for girls. Aggregation of externalizers did not affect engagement in prosocial influence or changes in this behavior over time. Interestingly, girls increased in prosocial influence over the summer, but boys did not. This basic group type effect was moderated by child type for the boys only. Mixed boys were especially likely to increase engagement in deviant talk if they were placed in a high-externalizing group, and mixed boys in lowexternalizing groups showed decreased deviant talk engagement. Non-externalizing and pure externalizing boys’ engagement in deviant talk increased in both low- and highexternalizing groups, but the increase was significant for both types of boys in highexternalizing groups. The same was not true for girls, as there were no three-way interactions for any child-group combination. Child Behavior Type and Group Composition Effects on Behavior Change

Residential Iatrogenic Effects 73 Whereas levels of deviant talk were broadly influenced by group type: high versus low externalizing membership, effects of group type on treatment effects were highly dependent on child type. Non-externalizing youth were affected most by group composition in behavior change analyses. Non-externalizing youth in high externalizing groups showed increased levels of problem behavior in analyses of total problem behavior, externalizing behavior, and aggression. Behaviors for non-externalizing boys in low versus high externalizing groups often did not differ until the end of treatment, but were headed that way over the course of the summer, suggesting a slow and steady influence of group membership on the behavior of these boys. For example, nonexternalizing boys in high-externalizing groups showed a more gradual increase in antisocial influence from Time 1 to Time 2 compared to mixed and externalizing boys in high-externalizing boys. Externalizers did not change behavior much over the summer, and were somewhat, but not as strongly influence by group membership. Externalizers in low-externalizing groups often showed a peak in behavior problems (externalizing, total problem behavior, aggression ) at Time 2, followed by a decline. This might indicate a period of elevated bullying to establish a dominant role in groups (non-externalizing), where dominance was possible for them. Most externalizing youth, however, did return almost to their initial levels for both externalizing behavior and total problem behavior, suggesting a temporary increase and no lasting harm from group membership. It is important to note that externalizers did not worsen when they were in groups with other externalizers, as would be predicted by the deviancy training hypothesis. They “worsened” when allowed the opportunity to dominate other less externalizing peers.

Residential Iatrogenic Effects 74 It was expected that mixed youth might show the greatest influence of group membership on behavior change because their deviant talk levels were most affected by group membership. Change in behavior, however, was not strongly influenced by group membership for mixed youth. Mixed youth did show some evidence of more problem behavior in high-externalizing groups, but only for externalizing and aggression, and only at Time 1. Thus, mixed youth sometimes showed early elevations in problem behaviors when they were with pure externalizers, but that group membership did not significantly influence how much they benefited from treatment. Boys’ total problem behavior showed signs of increase over the summer, but it was often qualified by child type and group type interactions. Parallel to findings for the boys, there was also a significant increase in total problem behavior for non-externalizing children, as well as a drop in total problem behavior for externalizing girls. However, externalizing girls were less affected by group membership. It is important to consider the low power to detect a three-way interaction, due to the small sample of girls in the present study. With respect to overall change and relationships with adults as measured by the CGI, non-externalizing boys were again most negatively affected by the presence of externalizing peers in clinical groups, although their behavior was still measured as “improved.” In contrast, mixed boys showed improvement in both low- and highexternalizing groups, but those in high-externalizing groups, showed more improvement, particularly for the global measure. For girls, the only significant finding was for child type. Externalizing girls improved less in peer relationships compared to nonexternalizing and mixed girls, regardless of group type. One important point to consider

Residential Iatrogenic Effects 75 is the high proportion of externalizing girls in the sample. Because there was not a true low-externalizing girls’ group to use as a reference point, it is understandable that there were no significant group effects. Differences between TRF and CGI findings are discussed in a later section. Gender Effects Boys and girls showed, at times, conflicting peer contagion effects. Previous research suggests that boys are more likely to engage in delinquent behaviors including gang membership (Gifford-Smith, Dodge, Dishion, & McCord, 2004). Furthermore, the development of male gangs seems to be independent, compared to female gangs, which tend to develop as an adjunct counterpart to boys’ groups (Decker, & Van Winkle, 1996). Thus, it is not surprising that deviant peer group influence may operate differently for boys and girls. To date, however, very little is known about gender differences in this area. On average, the boys’ groups were more heterogeneous. As demonstrated by the wider range of externalizing membership percentages. Using a 33% cutoff of the ratio of pure externalizers in a clinical group to classify a group as “high-externalizing” yielded significant gender differences. Of nine girls’ groups, 8 were identified as “highexternalizing” by this standard. For boys, 8 clinical groups met the high-externalizing classification, and 11 were low-externalizing at 33%. Because the boys groups were split at almost half and half for high-externalizing and non-externalizing, the cutoff was kept at 33%. For the girls, however, the cutoff ratio of pure externalizers was raised to 50%, which yielded 5 high- and 4 low-externalizing girls’ groups. Therefore, even in lowexternalizing groups, a cluster of externalizing girls was usually still present. Often,

Residential Iatrogenic Effects 76 nearly half of the girls in low-externalizing groups were externalizing, so I did not have a great high-low comparison for girls. I still found some influence of level of externalizing membership (which should be considered as moderate versus high rather than high versus low externalizing), even though the two levels were not as different as they were for the boys. I did not get child type by group type interactions, and this may have been because I had very few non-externalizing girls. It was the non-externalizing boys who were most strongly influenced by group membership. For both boys’ and girls’ groups, nonexternalizing youth were found to be most consistently influenced by externalizing group membership. For boys, only non-externalizing youth in high-externalizing groups were affected. For girls groups, non-externalizing youth in both high- and low-externalizing groups showed dampened treatment effects, but because of the highly homogeneous sample, we can conclude that all the girls’ groups were high in externalizing. Externalizing and non-externalizing girls’ groups were hardly different, but the line to distinguish the two types of groups had to be drawn somewhere. To that extent, the sample did not provide a good estimate of group-type influence, simply because there were no true low-externalizing girls’ groups. Perhaps in homogeneous girls’ groups, the norm was deviant behavior, so girls acted in accordance with the trend to avoid being the social misfit (Wright et al., 1986). For girls, group mattered for deviant talk and for total behavior problems, but for total behavior problems, there was no time x child x group interaction, only the two-way child by group interaction. It shows that non-externalizing girls were affected in ways similar to the boys, where mixed girls had a strange opposite effect and externalizing girls were not highly affected. According to data from the present study, externalizing

Residential Iatrogenic Effects 77 behavior and deviant talk are only correlated at r =.526, which is interesting given the concern for externalizers and deviant talk in the literature. Focusing on those who engage in deviant talk might be a better way to gauge group influence compared to focusing on externalizing youth by default. The present study suggests that the assumption that if a group of externalizing youth get together, they engage in deviant behaviors is not completely unwarranted. That is, perhaps it would be better to analyze externalizing youth directly and individually, rather than worry about their presence in a group. Initial results from the study suggest that group composition does matter. It influences escalation in deviant talk for both, but it only influences treatment effects for boys. However, because the overall levels of externalizing girls were so high, it is possible that the results are not complete. Obviously a larger sample of girls with a wider range of behavior problems would yield more definitive results. Treatment Effects & Iatrogenic Effects We find evidence of iatrogenic effects at Wediko not with externalizing youth, but with the non-externalizing youth forced to be in a group with externalizing youth. The literature suggests that placing non-externalizing youth in treatment groups with externalizing peers is not harmful, partially because externalizing youth are most likely to engage in externalizing behavior with their externalizing friends, rather than with internalizing or non-aggressive peers (Dishion, 2000). It is important to acknowledge, however, that studies of that nature often recruit truly non-externalizing, nonsymptomatic children to act as peer role models in the treatment setting. In clinical treatment programs, one could expect that all the children attending show at least some degree of problem behavior, or they wouldn’t be there. So why might non-aggressive

Residential Iatrogenic Effects 78 peer models not be harmed in past research, but non-aggressive youth in clinical treatment be affected as we observed? Perhaps iatrogenic effects depend on how welladjusted the child is and the amount of contact there is between the non- and highexternalizing youth over the course of treatment. In terms of adjustment, even the original “non-clinical” sample in the present study showed some degree of externalizing behaviors. Compared to the mixed and externalizing children in the population, however, their problem behavior seemed insignificant. In terms of exposure and contact, results from the present study suggest that it took the non-externalizing children all summer to become significantly more impaired in externalizing groups compared to their counterparts in the low-externalizing groups. Thus the process may be slow and gradual and require extended exposure to deviant peer models. Perhaps the non-externalizing children became the targets of bullying over the course of the summer, and their changes in aggressive behavior might be a result of self-defense in response to social rejection. Much of the previous research on iatrogenic effects that employed a non-symptomatic, non-externalizing sample did not expose them to high-externalizing peers for extended periods of time. Measures of Change Another point to consider is how the TRF and CGI measure change and which (if either) is more accurate. Different outcome measures in general often lead to difficult conclusions. Connor, Miller, Cunningham, and Melloni (2002) examined child improvement measures and treatment effects. In a program similar to Wediko, Connor et al. measured change of 87 children and adolescents with serious emotional disturbances. To measure psychopathology, Connor et al. used the Devereux Scale of Mental Disorder

Residential Iatrogenic Effects 79 (DSMD), a 110-item behavior rating scale derived by Naglieri, LeBuffe, and Pfeiffer (1994). Similar to the TRF, the DSMD is comprised of six narrow scales including conduct, attention/delinquency, anxiety, depression, autism, and acute problems as well as three composite scales (externalizing, internalizing, and critical pathology) and a total scale score. Connor et al. also used the Clinical Global Impressions Scale (CGI) and a verbal IQ scale. All measures were completed by the residential clinical staff team for target children. Previous research suggests that children with more pathological behavior show more positive outcomes in residential care (Hoagwood & Cunningham, 1992). In support of that research, Connor et al. found that children and adolescents with higher internalizing DSMD scores at admission to the treatment showed less psychopathology than youths with lower initial internalizing behaviors at the end. For externalizing behaviors, results were much less conclusive. Youths admitted to the program with elevated levels of externalizing behaviors appeared neither in the much improved or much worse groups (based on symptom change), which suggests that residential treatment may not be the most effective treatment option for highly externalizing youth. However, Connor et al. (2002) found that these youth improved based on staff impressions of change (CGI). Furthermore, Connor et al. found little agreement between DSMD assessments of improvement and clinical staff impressions of which children improved and got worse over the course of treatment. In the present research, there is sometimes evidence of increased behavior problems on the TRF, a measure much like the DSMD. This is often qualified by interactions, and rarely is there a steady linear increase in behavior problems. However, this pattern contrasts with the CGI which suggested general improvement, even for groups whose

Residential Iatrogenic Effects 80 treatment effects were dampened by externalizing group membership. The CGI is broader and captures all types of behavior, including adaptive ones. Also, having multiple raters for each child and averaging responses allows us to reach the “best estimate” that Connor et al. recommend. Connor et al. (2002) contrasted the strengths and weaknesses of measures like the TRF and impression measures of the CGI. These and other observations are relevant to the present study. The TRF is standardized and are more empirically based than other measures of change, like the CGI, which is a definite strength. However, the snapshot impressions of children staff provide at each of three assessments points may not well capture the changes children make. The TRF may not be sensitive to change because it pinpoints a child’s behavior at a particular point in time, rather than assessing change gradually over the summer. Further, raters may get stuck on an initial impression and then their ratings do not change much. More importantly, the TRF also only focused on behavior problems, while the CGI includes both problem and adaptive behaviors. It is important to acknowledge that prosocial gains are often the most important type of behavior change. For some children, learning valuable social skills and making friends can be very important, even if they are unable to significantly reduce their aggressive behavior. Although useful because it assesses many aspect of change, the CGI has its own weaknesses It is based entirely on staff impressions of change over a longer time period, which can invite wishful thinking. Staff may want to believe that children in the program are improving, and it is easy for those desires to come out on paper. Although different in the details, and different in the impressions they create about iatrogenesis, both measures support key findings in the present study.

Residential Iatrogenic Effects 81 While there were some differences in results across the CGI and TRF with respect to behavior change, both indicated that non-externalizing youth were most affected by which type of group they were assigned to. The TRF indicated that non-externalizing boys in externalizing groups show worse behavior over the summer, while the CGI indicated that the same boys just improved less compared to the non-externalizing boys in non-externalizing groups. Furthermore, with respect to externalizing boys, CGI results suggest that these boys were relatively unaffected by group composition. On the other hand, the TRF showed a slight “bullying” effect midsummer, with elevated problem behavior at Time 2. This effect was not captured by the CGI. The differences observed in results across the CGI and TRF leave us with questions. Does residential treatment work? Are youth harmed? Or are their treatment effects merely dampened? From these data, it is not totally clear that residential treatment works, but it is not totally clear that it truly harms. One possible exception is non-externalizing boys in externalizing groups. Yet even for these youth, CGI impressions of change are positive. It is important to note that measuring change is complex and results may vary depending on the measure used. Previous research at Wediko using a range of measures (observational, self-report, adult-report) and different theoretical approaches (syndromal, contextual) has shown that “change is multivariate and context-dependent, with children simultaneously showing both ‘improvement’ in some micro-contexts and ‘worsening’ in others” (Choukas-Bradley, Metcalfe, Wright, Zakriski, & Cardoos, 2009). Previous research using the TRF also suggested ‘worsening’ for externalizing and total problem behavior (Choukas-Bradley et al., 2009). In contrast, impressions of improvement are

Residential Iatrogenic Effects 82 positive. Choukas-Bradley et al. (2009) suggest that impressions of change on measures such as the CGI may be based on increases in prosocial behaviors that are not assessed by standardized measures such as the TRF. In fact, research by Choukas-Bradley, Banducci, Metcalfe, Wright, and Zakriski (2008) indicated discrepancies between staff ratings on impressions of improvement measures (CGI) and standardized assessments (TRF), which did not show change. Staff do report greater improvement for prosocial behavior on the CGI than decreases in problem behaviors, suggesting a tendency for staff to look for positive changes. It is also possible that in the compressed time period staff work with the children, it is difficult for counselors to report the change they see on bi-weekly behavior problem checklists, compared to the robust pre-post change reported by parents and teachers on the same type of measures. Although the TRF results suggest youth attending Wediko did not show improved behavior, daily field observations made by staff using the Wediko Behavior Observation System (WBOS), a context-reaction method of observation, indicate otherwise. The TRF may not be a particularly helpful measure in detecting behavioral changes that occur within the summer, because it does not examine behavior in context. Changes in problem behavior are narrow and context-specific on the WBOS, so these changes may go undetected by a summary measure like the TRF (Choukas-Bradley et al., 2008). As noted earlier, the TRF does not measure prosocial behavior, but the WBOS showed this to be an area in which youth showed many positive gains. Choukas-Bradley et al. (2009) indicate that different measures can reveal improvement or worsening for the same children. It is important to examine context-specific behavior, and to consider whether prosocial or antisocial behaviors are in question.

Residential Iatrogenic Effects 83 Limitations and Future Research There is a very limited body of research about residential treatment and aggression, so there are several possible directions for future research. The present study focused on differences in pure and mixed externalizers in a residential treatment program. Perhaps future research could further investigate these differences in long-term residential treatment, or in groups with different compositions. For example, what happens when mixed internalizing/externalizing youth are aggregated? Do they show elevated internalizing or externalizing behavior? The present study suggests that nonexternalizing youth are influenced by the presence of pure externalizers, but it would be interesting to observe how their behavior changes in a group composed of only nonexternalizing and mixed internalizing/externalizing youth. Furthermore, in the already incomplete literature about residential treatment, the literature focused on girls in treatment is extremely limited. Much of the present literature focuses on single-sex, male treatment programs. Even in co-ed residential treatment programs, girls are underrepresented in the literature. Although boys tend to populate deviant youth treatment programs more heavily, surely there is a larger antisocial female population than the current literature suggests. Perhaps future research could examine single-sex residential treatment, paying particular attention to girls’ programs. It is difficult to say whether treatment effects and behavior change have been adequately tested for girls in the present study due to the limited proportion of girls in the Wediko summer program, and the homogeneous sample of girls with externalizing problems. It would be useful to examine longer-term follow-up to see if iatrogenic effects show up over time. Parent data collected three to four months after treatment could be

Residential Iatrogenic Effects 84 examined in the present sample, but a longer term follow-up would also be important, as some studies have showed this delayed effect. Age may matter as a factor in vulnerability to iatrogenic effects, and this study did not adequately address this possible moderator. Age did not seem as important as gender in preliminary analyses, so gender became the focus. Age was covaried because externalizing groups were formed and differed in age. The body of literature might suggest that pre-adolescents are more susceptible to effects because that tends to be the time when selective friendships forms. It is likely that the effects continue into adolescence, but because primary friendships have typically developed by then, the susceptibility for iatrogenic effects is greater in pre-adolescence. The deviant talk measure on the peer assessment measure is only a two-item peer nomination measure. Past research uses dyadic observations to code occurrences of deviant talk and peer reinforcement. Such a measure is not feasible in a treatment program. Can kids accurately report on deviant talk? Our evidence suggests that they answered consistently on the peer nomination items for deviant talk, and that peer nominated deviant talk was sensibly related to variables like externalizing behavior that have been linked to observed measures of deviant talk. More research validating peer nomination methods for deviant talk assessment will be important for future treatment research. Also, the prosocial influence measure used only had one item, so further development of that scale will be important. Finally, although I demonstrated that non-externalizing youth showed increased deviant talk over time when in externalizing groups and they also showed increased behavior problems, a possible causal role of deviant talk exposure was only indirectly

Residential Iatrogenic Effects 85 demonstrated. Regression analyses for testing mediator effects would be useful in future analyses. Conclusions As predicted, externalizing youth engage in more deviant behavior and less prosocial behavior compared to mixed children. However, it is not as harmful to have clusters of externalizing children in residential treatment peer groups as was initially thought. The literature suggests that externalizers are the most harmed, but in fact, the non-externalizing youth were the most harmed. Externalizers alone are more problematic compared to mixed children, but the presence of externalizers in a group does not cause great harm, especially not to externalizers themselves. The presence of externalizers in clinical groups only affected treatment outcomes for non-externalizing boys, who showed increased problem behavior with improvement in interpersonal skills. Residential treatment programs should carefully monitor deviant talk within clinical groups, especially when there is a critical mass of externalizing youth in the group. The effects of deviant talk and how they unfold should be monitored especially closely for nonaggressive youth.

Residential Iatrogenic Effects 86 References Achenbach, T.M. (1991). Integrative guide for the 1991 CBCL/4-18, YSR, and TRF profiles. Burlington, VT: University of Vermont, Department of Psychiatry. Achenbach, T.M., & Rescorla, L.A. (2001). Manual for the ASEBA School-Age Forms & Profiles. Burlington: Research Center for Children, Youth, & Families, University of Vermont. Ang, R.P., & Hughes, J.N. (2002). Differential benefits of skills training with antisocial youth based on group composition: A meta-analysis investigation. School Psychology Review, 31, 164-185. Boxer, P., Guerra, N.G., Huesmann, L., & Morales, J. (2005). Proximal peer-level effects of a small-group selected prevention on aggression in elementary school children: An investigation of the peer contagion hypothesis. Journal of Abnormal Child Psychology, 33, 325-328. Bronfenbrenner, U. (1979). The ecology of human development: Experiments by nature and by design. Cambridge, MA: Harvard University Press. Buehler, R.E., Patterson, G.R., & Furniss, J.M. (1966). The reinforcement of behavior in institutional settings. Behavior Research and Therapy, 4, 157-167. Cairns, R.B., Cairns, B.D., Neckerman, H.J., Gest, S.D., & Gariepy, J.L. (1988). Social networks and aggressive behavior: Peer support or peer rejection? Developmental Psychology, 24, 815-823. Capaldi, D.M. (1992). Co-occurrence of conduct problems and depressive mood in early adolescent boys: II.A 2-year follow-up at 8th grade. Development and Psychopathology,4, 125–144.

Residential Iatrogenic Effects 87 Cardoos, S. (2006). Gender differences in deviancy training in a clinical setting. Unpublished honors thesis. Presented at the Connecticut College Annual Psychology Conference. Cardoos, S., Zakriski, A. L., Wright, J. C., & Parad, H. W. (2008). Deviant Talk in Residential Treatment: Individual and Group Influences. Poster proposal submitted to Society for Research on Child Development, Denver, CO. Cardoos, S.L., Zakriski, A.L., Wright, J,C,, & Parad, H.W. (2008b). Group treatment for aggressive and delinquent youth: Does peer interaction reinforce deviant behaviors? Brown University Child and Adolescent Behavior Letter, 24, 1-6. Cho, H., Hallfors, D.D., & Sanchez, V. (2005). Evaluation of a high school peer group intervention for at-risk youth. Journal of Abnormal Child Psychology, 33, 363374. Choukas-Bradley, S., Banducci, A. N., Metcalfe, L. A., Wright, J. C., & Zakriski, A. L. (2008). Reassessing the assessment of change: Disentangling the social interactional processes that mediate behavior change in at-risk youth. Poster presented at the Eastern Psychological Association, Boston, Massachusetts. Choukas-Bradley, S., Metcalfe, L. A., Wright, J. C., Zakriski, A. L., & Cardoos, S. (2009). Toward a Contextual Approach to the Study of Change: Linking Impressions, Standardized Assessments, and Behavior in Context. Poster presented at the Society for Research on Child Development, Denver, CO. Connor, D.F., Miller, K.P., Cunningham, J.A., & Melloni, R.H. (2002). What does getting better mean? Child improvement and measure of outcome in residential treatment. American Journal of Orthopsychiatry, 72(1), 110-117.

Residential Iatrogenic Effects 88 Decker, S.H., & Van Winkle, B (1996). Life in the gang: Family, friends, and violence. New York: Cambridge University Press. Dennis, M., Godley, S.H., Diamond, G., Tims, F.M., Babor, T., Donaldson, J., Liddle, H., Titus, J.C., Kaminer, Y., Webb, C., Hamilton, N., & Funk, R. (2004). The Cannabis Youth Treatment (CYS) study: Main findings from two randomized trials. Journal of Substance Abuse Treatment, 27, 197-213. Dishion, T.J. (2000). Cross-setting consistency in early adolescent psychopathology: Deviant friendships and problem behavior sequelae. Journal of Personality, 68, 1109-1126. Dishion, T.J., & Andrews, D.W. (1995). Preventing escalation in problem behaviors with high-risk adolescents: Immediate and 1-year outcomes. Journal of Consulting and Clinical Psychology, 63, 538-548. Dishion, T.J., Andrews, D.M., & Crosby, L. (1995). Antisocial boys and their friends in early adolescence: Relationship characteristics, quality, and interactional processes. Child Development, 66, 139-151. Dishion, T.J., & Dodge, K.A. (2005). Peer contagion in interventions for children and adolescents: Moving towards an understanding of the ecology and dynamics of change. Journal of Abnormal Child Psychology, 33, 395-400. Dishion, T.J., McCord, J., & Poulin, F. (1999). When interventions harm: Peer groups and problem behavior. American Psychologist, 54, 755-764. Dishion, T.J., Spracklen, K.M., & Patterson, G.R. (1996). Deviancy training in male adolescent friendships. Behavior Therapy, 27, 373-390.

Residential Iatrogenic Effects 89 Dodge, K.A., Dishion, T.J., & Lansford, J.E. (2006). Deviant peer influences in programs for youth: Problems and solutions. New York, NY: Guilford Press. Duncan, G.J., Boisjoly, K., Kremer, M., Levy, D.M., & Eccles, J. (2005). Peer effects in drug use and sex among college students. Journal of Abnormal Child Psychology, 33, 375-385. Espelage, D.L., Holt, M.K., Henkel, R.R. (2003). Examination of peer-group contextual effects on aggression during early adolescence. Child Development, 74, 205-220. Feldman, R.A. (1992). The St. Louis Experiment: Effective treatment of antisocial youths in prosocial peer groups. In J. McCord and R.E. Tremblay (Eds.), Preventing antisocial behavior: Interventions from birth to adolescence (pp. 233-252). New York: Guilford Press. Feldman, R.A., Caplinger, T.E., & Wodarski, J.S. (1983). The Saint Louis conundrum: The effective treatment of antisocial youth. Englewood Cliffs, NJ: Prentice-Hall. Frensch, K.M., & Cameron, G. (2002). Treatment of choice or a last resort? A review of residential mental health placements for children and youth. Child & Youth Care Forum. 31(5), 307-339. Gifford-Smith, M., Dodge, K.A., Dishion, T.J., & McCord, J. (2004). Peer influence in children and adolescents: Crossing the bridge from developmental to intervention science. Journal of Abnormal Child Psychology, 33, 255-265. Goodman, S. H., Lahey, B.B., Fielding, B., Dulcan, M., Narrow, W., & Reiger, D. (1997). Representativeness of clinical samples of youths with mental disorders: A preliminary population-based study. Journal of Abnormal Psychology, 106, 3-14.

Residential Iatrogenic Effects 90 Guy, W. (1976). Assessment manual for psychopharmacology. Washington, DC: U.S. Department of Health, Education, and Welfare. Hair, H.J. (2005).Outcomes for children and adolescents after residential treatment: A review of research from 1993 to 2003. Journal of Child and Family Studies. 14, 551-575. Hanish, L.D., Martin, C.L., Fabes, R.A., Leonard, S., & Herzog, M. (2005). Exposure to externalizing peers in early childhood: Homophily and peer contagion process. Journal of Abnormal Child Psychology, 33, 267-281. Hoagwood, K., & Cunningham, M. (1992). Outcomes of children with emotional disturbances in residential treatment for educational purposes. Journal of Child and Family Studies, 1, 129-140. Keenan, K., & Shaw, D. (1997). Developmental and social influences on young girls’ early problem behavior. Psychological Bulletin, 121, 95-113. Kellam, S. G., Brown, C. H., Rubin, B. R., & Ensminger,M. E. (1983). Paths leading to teenage psychiatric symptoms and substance use: Developmental epidemiological studies in Woodlawn. In S. R. Guze, F. J. Earns, and J. E. Barrett (Eds.), Childhood psychopathology and development (pp. 17–51). New York: Raven Press. Klein, M.W. (2006). Peer effects in naturally occurring groups: The case of street gangs. In K.A. Dodge, T.J. Dishion, and J.E. Lansford (Eds.), Deviant peer influences in programs for youth: Problems and solutions (234-250). New York, NY: Guilford Press.

Residential Iatrogenic Effects 91 Knorth, E.J., Klomp, M., Van den Bergh, P.M., & Noom, M.J. (2007). Aggressive adolescents in residential care: A selective review of treatment requirements and models. Adolescence, 42(167), 461-485. Kretsch, N.S., Olds, K.M., Wright, J.C., Zakriski, A.L., & Cardoos, S.L. (2008, March). Effects of person-group similarity on children’s social experiences during shortterm residential treatment. Poster presented at the annual meeting of the Eastern Psychological Association, Boston. Ladd, G.W., & Mize, J. (1983). A cognitive-social learning model of social-skill training. Psychological review, 90, 127-157. Lyons, J.S., Terry, P., Martinovich, Z., Peterson, J., & Bouska, B. (2001). Outcome trajectories for adolescents in residential treatment: A statewide evaluation. Journal of Child and Family Studies, 10, 333-345. Mager, W., Milich, R., Harris, M.J., & Howard, A. (2005). Intervention groups for adolescents with conduct problems: Is aggregation harmful or helpful? Journal of Abnormal Child Psychology, 33, 349-362. McCord, J. (1978). A thirty-year follow up of treatment effects. American Psychologist, 33, 284-289. McCord, J. (1992). The Cambridge-Somerville Study: A pioneering longitudinalexperimental study of delinquency prevention. In J. McCord & R.E. Tremblay (Eds.), Preventing antisocial behavior: Interventions from birth through adolescence (pp. 196-206). New York: Guilford Press.

Residential Iatrogenic Effects 92 McCurdy, B.L., & McIntyre, E.K. (2004). 'And what about residential...?' Reconceptualizing residential treatment as a stop-gap service for youth with emotional and behavioral disorders. Behavioral Interventions. 19, 137-158. Naglieri, J.A., LeBuffe, P.A., & Pfeiffer, S.I. (1994). Devereaux Scales of Mental Disorders. San Antonia, TX: Psychological Corporation. Poulin, F., Dishion, T.J., & Burraston, B. (2001). 3-year iatrogenic effects associated with aggregating high-risk adolescents in cognitive-behavioral preventive interventions. Applied Developmental Science, 5, 214-224. Poulin, F., Dishion, T.J., & Haas, E. (1999). The peer influence paradox: Friendship quality and deviancy training with male adolescent friendships. Merrill-Palmer Quarterly, 45, 42-57. Shapiro, J.P., Welker, C.J., Pierce, J.L. (1999). An evaluation of residential treatment for youth with mental health and delinquency-related problems. Residential Treatment for Children and Youth, 17, 33-48. Smith, E.P., Dumas, J.E., & Prinz, R. (2006). Prevention approaches to improve child and adolescent behavior and reduce deviant peer influence. In K.A. Dodge, T.J. Dishion, and J.E. Lansford (Eds.), Deviant peer influences in programs for youth: Problems and solutions (pp. 296-311). New York, NY: Guilford Press. Snyder, J., Brooker, M. Patrick, M.R., Snyder, A., Schrepferman, L., & Stoolmiller, M. (2003). Observed peer victimization during early elementary school: Continuity, growth, and relation to risk for child antisocial and depressive behavior. Child Development, 74, 1-18.

Residential Iatrogenic Effects 93 Snyder, J., Schrepferman, L., Oeser, J., Patterson, G., Stoolmiller, M., Johnson, K., & Snyder, A. (2005). Deviancy training and association with deviant peers in young children: Occurrence and contribution to early-onset conduct problems. Development and Psychopathology, 17, 397-413. Snyder, J., West, L., Stockemer, V., Givens, S., & Almquist-Parks, L. (1996). A social learning model of peer choice in the natural environment. Journal of Applied Developmental Psychology, 17, 215-237. Wright, J.C., Giammarino, M., & Parad, H. (1986). Social status in small groups: Individual-group similarity and the social “misfit.” Journal of Personality and Social Psychology, 50, 523-536. Zahn-Waxler, C. (1993). Warriors and worriers: Gender and psychopathology. Development and Psychopathology, 5, 79-89.