You are on page 1of 14

Clinical Psychological Science

http://cpx.sagepub.com/ The Buddy System: A 35-Year Follow-Up of Criminal Offenses


Clifford R. O'Donnell and Izaak L. Williams Clinical Psychological Science 2013 1: 54 originally published online 18 October 2012 DOI: 10.1177/2167702612456907 The online version of this article can be found at: http://cpx.sagepub.com/content/1/1/54

Published by:
http://www.sagepublications.com

On behalf of:

Association for Psychological Science

Additional services and information for Clinical Psychological Science can be found at: Email Alerts: http://cpx.sagepub.com/cgi/alerts Subscriptions: http://cpx.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav

>> Version of Record - Dec 14, 2012 OnlineFirst Version of Record - Oct 18, 2012 What is This?

Downloaded from cpx.sagepub.com at Alexandru Ioan Cuza on October 31, 2013

Empirical Article

The Buddy System: A 35-Year Follow-Up of Criminal Offenses


Clifford R. ODonnell and Izaak L. Williams
University of Hawaii, Honolulu

Clinical Psychological Science 1(1) 5466 The Author(s) 2013 Reprints and permission: sagepub.com/journalsPermissions.nav DOI: 10.1177/2167702612456907 http://cpx.sagepub.com

Abstract Court arrest records were obtained on 475 men and women who participated in a youth mentoring program or were randomly assigned to a no-treatment control group 35 years previously. Overall, 46.9% were arrested as adults. The proportion of adults arrested varied from 10% to 75% depending on gender, arrests prior to referral, and program participation. The effects of the program continued into the adult years: Among those with an arrest before program referral, significantly fewer participants were arrested as adults; however, among those without an arrest before the mentoring program, significantly more female participants were arrested as adults. In addition, arrest records were obtained for 236 adults who were listed in their civil and criminal cases. Arrest rates differed significantly by types of adult relationships and by proportion of drug arrests. Activity theory, based on activities and relationships, was used to interpret the findings. Keywords mentoring, delinquency, Buddy System, activity setting, peer network
Received 5/10/12; Revision accepted 7/9/12

The Buddy System was a youth mentoring program created solely to prevent juvenile delinquency. The roots of youth mentoring go back about 3,000 years, when Odysseus placed the care of his infant son, Telemachos, with a companion named Mentor (Butler, 1900/1944). The largest mentoring organization in the United States, Big Brothers Big Sisters, was founded in 1904 (Big Brothers Big Sisters, 2011). By 2008, there were several thousand mentoring programs with an estimated total of 3 million youth, six times the number from 10 years before (Rhodes, 2008). The specific goals of youth mentoring vary with the program, but all involve a relationship of a more experienced person to provide assistance or guidance to a child or adolescent. Reviews of mentoring studies have shown modest positive effects across a variety of outcome measures, including academic performance, attitudes, health behaviors, improved relationships, delinquency, and drug use (DuBois, Portillo, Rhodes, Silverthorn, & Valentine, 2011; Rhodes, 2008; Rhodes & DuBois, 2008; Tolan, Henry, Schoeny, & Bass, 2008; Wheeler, Keller, & DuBois, 2010), but there have not been any long-term studies on the effects of youth mentoring on adult criminal behavior. A meta-analysis of 73 independent evaluations noted the need for evaluations of youth offending and longer-term follow-up studies (DuBois et al., 2011). The purpose of the current study is to assess the effect of the Buddy System on adult criminal offenses 35 years later. The Buddy System included 553 males and females, 11 to 17

years of age (M = 13.5, SD = 1.01), who lived in one of two lower-income areas of Hawaii (one rural, the other urban). Although their ethnic/racial backgrounds included a mixture of Hawaiian, Filipino, Japanese, Chinese, and Caucasian, culturally they were all local. As is common in Hawaii, especially in these neighborhoods, their culture was not based on their ethnic/racial mixtures but on the local culture that was formed in Hawaii from the interactions and intermarriage among the many different groups. Pidgin is the language of local culture and the marker of local identity. All of these adolescents had multiple ethnic/racial backgrounds, but all spoke Pidgin. Given the complexities of the many ethnic/racial combinations and the reality that all were local, no attempt was made to obtain the frequency of each background or to subsequently analyze by ethnicity/racial background. Any attempt to do so would have been inaccurate and culturally insensitive (for the history and characteristics of local, see Okamura, 1994; Reinecke, 1969; Stannard, 2005). Intervention in the program was based on the triadic model developed in the classic work of Tharp and Wetzel (1969). In their model, consultants supervised mediators who had preexisting relationships with adolescents considered at high risk for
Corresponding Author: Clifford R. ODonnell, Department of Psychology, University of Hawaii, 2530 Dole St., Honolulu, HI 96822-2216 E-mail: cliffo@hawaii.edu

The Buddy System delinquency; the consultants were professionals, and the adults were often teachers or parents. The triadic model allows a small number of consultants to supervise a larger number of mediators and thereby intervene with a larger number of people. The mediators were trained by the consultants to use behavior modification intervention to improve the behavior of the adolescents. In the Buddy System, the consultants were advanced graduate students in clinical psychology who were supervised by university faculty and specialists. The mediators were adults (buddies) who were recruited through advertisements in the newspapers from the same lower-income neighborhoods as the adolescents. Mediators were hired and trained to mentor three adolescents each, and they were paid monthly for completing specific tasksincluding training, meeting with their assigned mentees every week, and submitting data to the consultants. The buddies were not randomly assigned to their mentees. Instead, assignment preferences of parents, adolescents, and buddies, regarding geographical proximity, gender, or mutual interests were accommodated whenever possible. The adolescents were individually referredprimarily by school officials but also by parents, neighbors, and social welfare, police, and court personnelfor specific behaviors, such as aggression, truancy, and poor academic performance. These behaviors were considered to place the adolescents at higher risk for delinquency and, eventually, adult criminal offenses. Signed parental permission was required for the assignment of an adolescent to a buddy. Initial attempts to contact parents by mail or phone had limited success, so the buddies made contact through visits to explain the program and obtain permission from those who wanted to participate. Many more adolescents were referred than could be admitted to the program. This excess of referrals permitted the random assignment of the adolescents to either the program or a no-treatment control group and resulted in a greater number of participants than controls. With a faculty supervisor, the consultants assigned referrals to the mentoring program or the control group individually using a random number table. Data were obtained from school and court records for those in the control group to assess program effectiveness, but no contact was made with these adolescents. The purpose of the control group was to evaluate the effects of intervention by comparing the mentoring program with a group that did not receive any planned intervention. Referrals were made at the beginning of each of the 3 years of the program and sometimes within the years. Some adolescents were referred more than oncefor instance, when adolescents assigned to the control group were referred the next year, when contact with parents was initially unsuccessful, when parents who first declined permission changed their minds, or when agency personnel continued to want an adolescent assigned to a buddy. Although all of those referred met the same criteria and participation was assigned randomly, multiple referrals could affect composition of the participation and control groups. Specifically, multiple referrals raised the possibility that a larger number of adolescents assigned to

55 buddies were at higher risk than those assigned to the control group. To assess this possibility, the arrest rate of participants referred only once was compared to the rate of participants referred more than once. The minor difference was not significant (Z = 0.48; ODonnell, Lygate, & Fo, 1979). Buddies had a small monthly allowance to spend on fun activities with their mentees. Examples of the activities, as selected by the adolescents with the consent of their mentors, included walks in Waikiki, going to movies, enjoying food treats, and surfing. For some of the participants, the activities were contingent on improvement in their referred behaviors, while for others the activities were noncontingent. These groups were created by individual random assignment using a random number table. When the contingency condition was shown to result in greater improvement, activities were made contingent for all of the participants. Comparisons then focused on differences between participants and controls. In these comparisons, every adolescent who ever participated in the mentoring program was considered in the participant group. These adolescents included those who participated in more than 1 year, as well as any who left the program during any year. The purpose of including all who participated for any length of time was to be sure that the participant group was not biased by a greater proportion of motivated, cooperative adolescents (for a more detailed description of the Buddy System, see ODonnell, 1992). The results were clear: When fun activities with the buddies were contingent on behavior, improvement on referred behaviorsas assessed by official school records and data submitted by the buddiesoccurred significantly more frequently than when the activities were noncontingent or when youth were randomly assigned to the no-treatment control group. The Buddy System successfully improved the behaviors for which the adolescents were referred (Fo & ODonnell, 1974). The expectation was that improvement on the referred behaviors would help to prevent delinquency, but these results were more complex. Over 3 years, those who participated in the Buddy System had a lower arrest rate than did those in the control group but only if they had an arrest in the year prior to referral (56.0% and 78.3%, respectively). The opposite was true for those without a prior arrest (22.5% and 16.4%, respectively). In other words, participation significantly decreased arrest rates for those previously arrested in the year before but significantly increased rates for those without a prior arrest in the year before referral (ODonnell et al., 1979). All of the arrests excluded juvenile status offenses. Considering only the arrests the year before referral as prior arrests allowed the periods to be equated for those who were referred in different years. The results supported preliminary findings that first identified the possibility of an iatrogenic effect (Fo & ODonnell, 1975) and were interpreted as a peer network effect, created during the many opportunities for youth to meet one another and form friendships during participation. These effects of youth with different levels of risk interacting with one another have been supported in two studies of

56 childrens aggression and adolescent substance abuse (Boxer, Guerra, Huesmann, & Morales, 2005; Valente et al., 2007). In both studies, the behavior of the higher-risk youth improved following the interaction but was worse for those at lower risk. A similar effect was found for youth interacting with teenage peer mentors (Karcher, Davidson, Rhodes, & Herrera, 2010). Reviews of intervention and delinquency studies since the Buddy System findings show strong evidence of the peer network effect. Diversion and delinquency prevention programs that offered social services to youth, arranged visits to adult prisons, or provided alternatives to gang membership typically showed either no effect or an increase in delinquency (McCord, 2003; ODonnell, Manos, & Chesney-Lind, 1987). Adverse effects also were found in a review of studies that grouped together high-risk youth for skill training (Arnold & Hughes, 1999). These findings are consistent with studies showing that most delinquent behavior occurs in groups and that the best predictor of delinquency is participation in activities with antisocial peers, delinquent friends, or youth gangs (ODonnell, 2005). Interest and recognition of the importance of the peer network effect have grown in recent years (e.g., Akers, 1998; Dishion & Tipsord, 2011; Dodge, Dishion, & Lansford, 2006) with studies detailing how changes in delinquency among adolescents depended on changes in the delinquency of their friends (Brendgen, Vitaro, & Bukowski, 2000; Vitaro, Brendgen, & Tremblay, 2001; Warr 1998) and how high-risk peers may encourage and support deviant behavior, increase the seriousness of violent behavior, and facilitate carrying weapons (Capaldi, Dishion, Stoolmiller, & Yoerger, 2001; Dijkstra et al., 2010; Dishion, McCord, & Poulin, 1999; Dishion, Vronneau, & Myers, 2010; Dodge, Greenberg, & Malone, 2008; Thornberry, 1998). The implications of the peer network effect extend beyond programs designed for adolescent peer groups and to the social systems in which they participate. Especially important are the implications for the juvenile justice, child welfare, child mental health, and education systems, which often facilitate contact with high-risk adolescents (Dodge et al., 2006; Dodge & Sherrill, 2006; Lipsey, 2006; Shapiro, Smith, Malone, & Collaro, 2010). A review of peer effects in the juvenile justice system concluded that institutional placement should be avoided whenever possible (Greenwood, 2006). Recent studies support that conclusion. One study, matching youth on predictors of recidivism, found higher recidivism rates for youth released from residential placements than for those released from community placements. The results were attributed to deviancy training in the high-density residential placements (Shapiro et al., 2010). These results were confirmed by a 20-year longitudinal study of a low socioeconomic cohort of 779 boys. Intervention by the juvenile justice system was more likely if the boys were hyperactive, had deviant friends, or were poorly supervised by their parents. Any intervention of the juvenile justice system greatly increased adult criminal offenses, with placement in a

ODonnell, Williams juvenile institution having the most negative effect (Gatti, Tremblay, & Vitaro, 2009). This surge of studies related to possible iatrogenic effects among adolescent peers raises the question of whether the effects continue after the adolescents become adults. These studies have not followed youth much beyond their adolescence, but some indications can be found among earlier longitudinal studies, including two classic studies with 30-year follow-ups. Boys who were referred to a child guidance clinic for antisocial behavior had higher rates of arrest and imprisonment as adults than did those in a matched control group who were not referred (Robins, 1966). Also, a greater proportion of boys who participated in the Cambridge-Somerville program, which provided a variety of counseling and social services to prevent delinquency, was convicted of more than one criminal offense as adults when compared to a matched, randomly assigned control group (McCord, 1978). The percentage of adults with criminal offenses in these studies (72% and 66%, respectively) illustrates the persistence of antisocial behavior into the adult years. Data on the adult years of the 500 male delinquents from the well-known study by Glueck and Glueck (1950) were provided by Laub and Sampson (2003). By age 24, 84% had been arrested, a rate that increased to 94% by their 50th birthday (76% for serious crimes). Although arrest rates declined with age, 23% were still arrested in their 50s and 12% in their 60s (21% for serious crimes after the age of 40). Considerable variation in desistence rates was found when 52 of these men were located and interviewed to gather information about how life circumstances were related to changes in their arrest rates. Controlling for age and other variables, desistence was related to transitions in life circumstances, such as marriage, employment, and military service. The largest effect was found for marriage, with an estimated 40% reduction in criminal offenses, even when controlling for the propensity to be married (Laub & Sampson, 2003). Marriage was also shown to affect offending in a Dutch cohort of male and female offenders (Blokland & Nieuwbeerta, 2005). The conviction rate declined 27% when low-rate offenders were married, and it increased 44% when they were separated. In contrast, high-rate offenders were less affected by changes in life circumstances. Although the conviction rate was higher for men than women, the marriage effect was similar for both (Petras, Nieuwbeeta, & Piquero, 2010). Other studies have shown that the marriage effect depends on changes in peer networks and the criminal activity of the spouse or romantic partner. A dramatic decline in time spent with friends and delinquent peers was reported for males and females following marriage, leading to a decline in criminal activity (Warr, 1998). In a study of a birth cohort of 983 men and women 18 to 21 years old, the rate of offending depended on whether cohort members were involved with a partner and if the partner was deviant. For both men and women, the highest rates of offending occurred for those with a deviant partner

The Buddy System and the lowest rates for those with a nondeviant partner (Woodward, Fergusson, & Horwood, 2002). These results suggest that having a partner can affect the rate of offending for better or worse among young adults. The importance of romantic partners on adult offending was supported in a longitudinal study of young adult couples (Simons, Stewart, Gordon, Conger, & Elder, 2002). For both men and women, adult networks with deviant members increased the probability of delinquents engaging in crime as adults. Men, however, were less likely to have deviant affiliations if they had a positive relationship with their romantic partner. This positive relationship had the indirect effect of reducing their chances for adult criminal behavior by affecting their adult networks. The strongest predictor of adult crime for women was an antisocial romantic partner. Their chance of having such a partner was increased if they associated with deviant peers as adolescents. In addition, while only adult deviant affiliations mattered for men, adolescent association with deviant peers was related to adult crime for women, independent of their partners and adult networks. These results suggest that the effect of adolescent peer networks continues into at least the early adult years for women, while adult networks are most predictive of adult crime for men. All of the studies reviewed have implications for assessing the adult criminal offenses of youth in the Buddy System and the control group. However, none of these studies assessed the long-term effect of youth mentoring programs on adult criminal offenses and, most importantly, on the possible continuation of an iatrogenic effect into the adult years. The question is whether the effects of youth mentoring on delinquency, positive and negative, continue to affect adult criminal activity. The Buddy System provides a unique opportunity to address this question because (a) a reasonably large sample of highrisk youth (n = 553), males and females, were randomly assigned to either a mentoring program or a no-treatment control group; (b) a positive effect on arrests was found for those arrested prior to referral and a negative effect for those not previously arrested; (c) sufficient time, 35 years, has elapsed to allow the long-term effects on adult criminal offenses to be assessed; and (d) the Hawaii State Judiciary has made information on court cases, criminal and civil, publicly available online. We expect the effects of the Buddy System, positive and negative, to be found in the analysis of adult offenses. Also, just as adolescent relationships influence arrest rates, adult relationships are expected to influence adult arrest rates. The studies reviewed showed that the types of adult relationships could influence adult criminal offenses positively or negatively. Specifically, these studies showed that marriage decreases criminal offending, especially for men; that male friendships increase criminal offending; and that the effect of romantic relationships depends on the deviancy of the partner (Blokland & Nieuwbeerta, 2005; Laub & Sampson, 2003; Petras et al., 2010; Simons et al., 2002; Warr, 1998; Woodward et al., 2002).

57 However, none of these studies on adult relationships assessed program effects or whether positive or negative program effects were influenced by these relationships. Therefore, it is important to examine adult relationships, especially those that are most likely to influence adult offenses, in this follow-up study. The criminal and civil records from the Hawaii State Judiciary lists the names of those involved in the cases, as well as the court documents for each case. The type of case and related court documents provide information on the nature of the relationships in question, such as spouse, romantic partner, friend, family member, or business associate. It is expected that these types of important relationships, as documented by court records, are most likely to influence adult offenses, especially relationships of male friendships and marriage. Specifically, we hypothesize the following: Hypothesis 1: Fewer of those who participated in the mentoring program and were arrested in any year prior to their referral to the Buddy System will have been arrested as adults than those in the control group (positive Buddy System effect). Hypothesis 2: More of those who participated in the mentoring program and did not have any arrests prior to referral will have been arrested as adults than those in the control group (iatrogenic Buddy System effect). Hypothesis 3: More of those with relationships to adults who were arrested will also have been arrested, with the highest proportion of arrests from those in male friendships and the lowest proportion from those in marriages. These relationship effects can then be examined to see if they contribute to an interpretation of the effects of the first and second hypotheses.

Method
Data collection on the adult criminal arrests of the participants in the Buddy System and the control group began in 2008, 35 years after the program, and it continued to 2010. In 2010 and 2011, data were obtained on the adult criminal arrests of those listed in the criminal and civil cases (divorces, temporary restraining orders, child support, lawsuits, motor vehicle torts, estate matters, and foreclosures). The Institutional Review Board approved an exemption for this study because all of the data were publicly available and no contact was made with any of the participants.

Procedure
All of the records of the Buddy System existed in the form of handwritten or typed documents. They were preserved in locked file cabinets for 35 years. From these records, a computer file was constructed of the 553 names, indicating gender,

58 age, schools, grade levels, neighborhoods, juvenile court records (if any), and whether they were in the mentoring program or the control group. Next, it was important to know if these adults lived at least part of their adult lives in Hawaii because the Hawaii State Judiciary reports records of arrests only for residents in Hawaii. Arrests out of state, if any, would not appear on these records. Publicly available records with information on age and addresses also were helpful to distinguish among people with the same name to be sure that the Judiciary records were of the correct people. Therefore, publicly available online databases were checked and cross-checked for information on the names of the 553 adults. These databases included people searches (Pipi, Peek You, Wink, Zabasearch, and Intellius), telephone and address directories (Polk City Directory, Switchboard, and White Pages), school alumni directories (Reunion and Classmates), and social networks (Myspace and Facebook). In some cases, it was necessary to check different versions of first names, as when Cathy was listed as Catherine or when Bob was Robert. A wealth of information was retrieved, often including birth dates, current and previous addresses, schools attended, phone numbers, marital status, and names of family members. Not all of this information was found for each name, so whenever there was any doubt if the record was for the correct person, copies of the court dockets were examined in person in the Court Records Archive. Sufficient information was found to correctly identify 476 of the 553 names and verify that they lived at least part of their adult lives in Hawaii, for a successful follow-up rate of 86.1%. As expected, this rate varied by gender. For men, the rate was 93.6%, including 94.7% of the participants and 92.1% of the controls. For women, the rate was 72.4%, including 78.3% of the participants and 61.2% of the controls. Of the 476 people, 7 were deceased, 1 before her 18th birthday. The 6 others3 male participants, 1 female participant, and 2 female controlsall died during their adult years, from their early 20s to their mid-40s, so adult criminal arrests were possible for them. With this information, the publicly available online database of the Hawaii State Judiciary was accessed first for the criminal arrests of the 475 remaining names, including listed aliases, and then for the names and type of relationships of other adults listed in the criminal and civil cases. The names of 236 other adults were found. These names were then processed for the records of their criminal arrests. A national criminal offense database (Criminal Searches) was also checked for out-of-state arrests, but none were found.

ODonnell, Williams group was composed of 139 males (77.2%) and 41 females (22.8%). Of the participants, 51 had been arrested prior to their referral to the Buddy System (17.3%): 40 males (20.5% of the male participants) and 11 females (11.0% of the female participants). Of the corresponding 180 controls, 24 were arrested prior to referral (13.3%): 22 males (15.8% of the male controls) and 2 females (4.9% of the female controls). All of the adults were now 46 to 52 years old (M = 48.5, SD = 1.01). Their ages depended on their age at referral and their year of referral. Names of other adults listed in the civil and criminal court cases were found for about half of the 475 adults (49.7%). Of those listed, 19.1% were female and 80.9% were male, reflecting the greater number of males in this study and the larger number with court records. The proportions from participants (63.6%) and controls (36.4%) closely matched those in the above sample. By examining the type of cases and court documents, it was possible to identify the type of relationships between 94.5% of the study sample (participants and controls) and those listed in the court records. All of these relationships involved spouses (52.5%), romantic partners (22.5%), friendships between males (13.6%), or family (5.9%). The relationships for the remaining 5.5% could not be ascertained.

Results
The Hawaii criminal court records were used to compute arrest rates, defined as the proportion of adults who had at least one arrest for a criminal offense. These data showed that 223 of the 475 youths were arrested as adults, yielding an arrest rate of 46.9%, with significantly higher rates for men than for women (56.9% and 23.4%), Z = 6.68, p < .001. Almost two-thirds of these arrests (65%) were for serious offenses (felonies) as defined by Loeber, Farrington, and Waschbusch (1998). The rate of arrests varied between 10% and 75% depending on gender, arrests prior to referral, and participation in the Buddy System. The rate of serious arrests followed the same pattern without any significant differences from all arrest rates. The number of arrests and type of offense for each person who was arrested were examined. The only significant difference for the number of arrests occurred between the mean number of arrests for men and women with an arrest, 2.93 and 1.67, respectively, t(221) = 2.99, p = .003. Men and women also differed significantly on type of offense, 2(3) = 10.13, p = .017. Men had proportionally more offenses against a person than women, 52.3% and 33.3%, respectively (Z = 2.69, p = .007), whereas women had proportionally more drugrelated offenses, 28.1% and 12.8% (Z = 3.10, p = .002). The proportions for property offenses (25.1% for men and 28.1% for women) and for other offenses (9.8% and 10.5%, respectively) were not significantly different. An analysis of each of the three hypotheses is presented next, followed by an examination of the relationship findings with the other significant effects.

Samples
The sample of 475 adults consisted of 295 participants (62.1%) in the Buddy System and 180 (37.9%) from the randomly assigned no-treatment control group. The participants included 195 males (66.1%) and 100 females (33.9%), and the control

The Buddy System The results were analyzed to assess each hypothesis. To assess the first hypothesis, arrest rates were analyzed for Buddy System participants and controls with an arrest prior to referral. With arrest rates based on proportions, a one-tailed Z test of proportions was used to analyze the predicted difference. This analysis supported the first hypothesis: The arrest rate for participants with an arrest prior to referral (54.9%) was significantly lower than the corresponding rate for controls (75.0%; Z = 1.67, p = .048). This result is consistent with program results 35 years ago; that is, the Buddy System lowered arrest rates for participants with an arrest prior to referral compared to the controls. Now as then, the small sample of females with an arrest prior to referral precluded meaningful comparisons of these arrests by gender. To assess the second hypothesis, arrest rates were analyzed for Buddy System participants and controls without an arrest prior to referral. First a chi-square analysis was used to see if the arrest rates for these participants and controls varied by gender. This analysis showed a significant difference, 2(3) = 10.63, p = .014. Therefore, one-tailed Z tests of proportions were used separately for females and males to determine the source of the predicted effect. The second hypothesis was supported by the significantly higher rate for female participants without an arrest prior to referral (29.2%) than for the corresponding controls (10.3%; Z = 2.32, p = .01). Contrary to prediction, the corresponding rates for males were virtually identical (53.5% and 54.7%), and the difference was not significant. This result is consistent with the previous program results only for females, with higher arrest rates for participants without an arrest prior to referral compared to the controls. The significant effects for the first and second hypotheses are presented in Table 1. The results for the first and second hypotheses were examined to see if the pattern of results could be explained by regression toward the mean. For regression toward the mean to be a viable explanation, the arrest rate should decrease for the group with a greater proportion of those with prior arrests and increase for the group with a greater proportion of those without prior arrests; that is, as one rate goes down, the other should go up. These related rates should show regression by moving in opposite directions. Female controls without an arrest prior to referral were a greater proportion of the total sample of female controls (95.1%) than were the corresponding female participants (89.0%). Therefore, any regression toward the mean should result in higher arrest rates for these female controls than the corresponding participants. Instead, the reverse occurred: The arrest rate of these female participants was almost triple the corresponding rate for the female controls. The proportions departed from, rather than moved toward, the mean. Also, the corresponding proportions for males (79.5% and 84.2%, respectively) produced virtually identical arrest rates, rather than regression toward the mean. In the analysis of the first hypothesis, the participant group had a somewhat greater proportion of those with an arrest prior to referral when compared to the controls (17.3% and
Table 1. Adult Arrest Rates by Gender, Participation, and Arrest Prior to Referral Group Total sample by gender Males Females Total Participants and controls without PA by gender Female participants Female controls Male participants Male controls With PA Participants Controls n 334 141 475 %a 56.9 23.4 46.9 2(3)

59

Z 6.68b

10.63c 2.32b

89 39 155 117 51 24

29.2 10.3 53.5 54.7 54.9 75.0

1.67b

Note: PA = prior arrest. The mean difference between participants and controls was not significant. In addition, the small sample of females with an arrest prior to referral precluded meaningful comparisons of these arrests by gender. a Indicates percentage of n arrested as adults. b Based on comparison with the percentage in the row below. c Based on comparison of the percentages in the indented rows below.

13.3%, respectively), but a Z test showed that this difference was not significant (Z = 0.44). Even more important, it is not plausible that the rates of these groups regressed toward the mean while the related rates of females without a prior arrest were in the opposite direction (i.e., away from the mean) and the rates for males without a prior arrest showed no regression. Therefore, a regression toward the mean interpretation for the results of the first two hypotheses was rejected. The purpose of the third hypothesis is to assess the potential influence of adult relationships on adult arrest rates and to contribute to our understanding of the effects of the Buddy System in the adult years of the participants and controls. The analysis of this hypothesis showed that of those with others listed in their civil or criminal court records, 44.1% of the listed relationships had at least one arrest for a criminal offense, while the remaining 55.9% did not have any arrests. As predicted, arrest rates were higher for those with relationships with arrests than for those with relationships without arrests (76.9% and 49.2%, respectively; Z = 4.34, p < .001). In addition, the arrest rates for those with relationships differed significantly by type of relationship, 2(3) = 22.98, p < .001. The sources of this difference were the higher rates for male friendships (90.6%), compared to the other relationships: romantic partners (56.6%; Z = 3.30, p = .001); marriage relationships (44.4%; Z = 4.67, p < .001); and family relationships (42.9%; Z = 3.49, p < .001). The arrest rates for crimes committed together also differed significantly by type of relationship, 2(3) = 25.84, p < .001. The source of this difference was the higher rates for male friendships (58.6%) and romantic partners (53.3%) compared to marriage relationships (10.9%), Z = 4.66 and 4.27, respectively, p < .001.

60 These data support the third hypothesis of the potential influence of adult relationships on arrest rates, especially for male friendships and marriages, and crimes committed together. Married couples were about five times less likely to be arrested together for the same crime as were either male friends or romantic partners. The significant effects for the third hypothesis are presented in Table 2.

ODonnell, Williams rate was about four times higher than the corresponding females without any relationships listed in their court records. With these findings indicating a high association with relationship arrest rates, the number and type of offenses of their relationships were examined. The mean number of arrests for participants relationships (5.93) was significantly higher than the corresponding mean for controls (1.85), t(39) = 2.29, p = .028. Also, a chi-square analysis indicated that there were significant differences among the types of relationship offenses, 2(3) = 20.31, p < .001. The source of this effect was the significant difference in drug arrests (33.7% and 0% for the relationships of participants and controls, respectively), Z = 3.39, p < .001. These findings, presented in Table 4, mean that the relationships of these female participants have a significantly higher average number of arrests than do the relationships of the corresponding controls, with about one-third of their arrests drug related. A high proportion of these female participants were arrested for offenses committed with their listed relationships (9 of 17, or 52.9%). Of the multiple arrests for offenses committed together, 19.1% were drug related (9 of 47). Corresponding numbers for controls were too low for meaningful statistical comparison, but none of the corresponding control arrests were committed together. Compared to the controls, these participants had relationships with persons who were much more likely to be arrested, to be arrested much more frequently, and to have a high proportion of their arrests related to drugs. In contrast, the proportion of drug arrests for the relationships of the corresponding males (22.0% and 25.0%, respectively) did not differ

Examination of significant effects and relationship data


The data on adult relationships provided an opportunity to examine the significant effect of the higher arrest rates of female participants without an arrest prior to referral. A chisquare analysis indicated that there were significant differences in arrest rates among these participants and corresponding controls with and without court records listing relationships, 2(3) = 27.95, p < .001. Z tests showed three sources of this significant effect: The arrest rate of those with these relationships was higher than those without these relationships (47.6% and 11.6%, respectively), Z = 4.52, p < .001. Also, the corresponding rate for participant females (58.6%) was significantly higher than that for controls (23.1%) and that for female participants without relationships listed (15%), Z = 2.13 and 4.25, p = .033 and p < .001, respectively. These findings, presented in Table 3, mean that not only were arrest rates significantly higher for female participants without a prior arrest, but so were the arrest rates of their relationships (more than double the rate for control relationships). Moreover, their arrest

Table 2. Adult Arrest Rates: Relationships With and Without Arrests and Type of Relationship Relationships with and without an arrest Adult arrest rate If relationships have an arrest If relationships dont have an arrest Arrest rates by type of relationship Male friendships Romantic partners Marriage Family Unknown Arrest rates for offenses committed together by type Marriage Male friendships Romantic partners Family Unknown Total
a b

n 104 132 32 53 124 14 13 55 29 30 6 6 126

%a 76.9 49.2 90.6 56.6 44.4 42.9 46.2 10.9 58.6 53.3 33.3 0.0 32.5

2(3)

Z 4.34

22.98b 3.30c 4.67c 3.49c 3.25c

25.84b

4.66d 4.27d

Indicates percentage of n arrested as adults. Based on comparison of the percentages in the indented rows below. c Based on comparison with male friendships. d Based on comparison with marriage.

The Buddy System


Table 3. Arrest Rates of Females Without Prior Arrest by Relationship Records Group Arrest rates by relationship records Participants and controls with records Participants and controls without records Controls with records Participants with records Participants without records n 42 86 13 29 60 %a 47.6 11.6 23.1 58.6 15.0 2.13c 4.25c 2(3) 27.95
b

61 trend for their rate to also be higher than the 27 corresponding participants (51.9%), Z = 1.71, p = .087. This result showed that this smaller sample of controls, with an arrest prior to referral, were the only group with a higher arrest rate for those without relationships listed in their court records. The arrest rate for all other groups was lower for those without relationship records. This finding supports the interpretation that contact with lower-risk peers, as occurred with participants, but not controls, may have lowered the arrest rate for those with a prior arrest. With the relationships of females examined, comparisons were made of the relationships of men with no arrests (participants and controls combined). The 68 men were significantly more likely to have relationships with others who also were arrest-free than with others with an arrest (83.8% and 16.2%, respectively), Z = 7.88, p < .001. If their court records listed only females (n = 129), they were significantly more likely to be arrest-free than if only males were listed (n = 27), 75.2% and 3.7%, respectively, Z = 13.45, p < .001, or if both males and females were listed (n = 35), 42.9%, Z = 3.64, p < .001. These findings indicate that being arrest-free is more likely for men (a) if the others in their relationships also have not been arrested and (b) if the others are female.

Z 4.52c

Note: The difference between participants and controls without relationships listed in their court records was not significant. a Indicates percentage of n arrested as adults. b Based on comparison of the percentages. c Based on comparison with the percentage in the row below.

significantly, nor did their arrest rates or average number of arrests. The gender differences in the rates and proportion of drug arrests among the relationships of these females and males contribute to our understanding of (a) the difference of the significantly higher arrest rate for these participant females and (b) the lack of a difference for corresponding males. The overall arrest rates and the proportion of drug arrests were significantly higher for the relationships of these female participants than for the corresponding female controls, but these differences did not occur with the corresponding males. Consistent with the third hypothesis, these results support the potential influence of adult relationships on the higher arrest rate for these female participants. Similar analyses were conducted for the significantly lower rate for participants with an arrest prior to referral found in the analysis of the second hypothesis. No significant results were found for those with relationships or for type of offense. In contrast, the 11 controls without court records of adult relationships had a significantly higher arrest rate than the corresponding rate for all of the 475 participants and controls (81.8% and 32.2%, respectively), Z = 3.39, p < .001, and a

Discussion
The results of this study indicated the continuation of positive and negative effects of the Buddy System 35 years later, supporting the first and second hypotheses. The first hypothesis predicted that Buddy System participants with an arrest prior to referral would have a lower arrest rate than that of the corresponding controls. The respective rates of 54.9% and 75% differed significantly. The second hypothesis predicted that participants without an arrest prior to referral would have a higher arrest rate than would the corresponding controls. This hypothesis was supported only for females. These participant females had a significantly higher arrest rate than did the corresponding controls, 29.2% and 10.3%, respectively. The difference for corresponding males was not significant. Examination of these data indicated that these results were not due to regression toward the mean. The third hypothesison

Table 4. Relationship Arrests and Offense Type by Participation for Females Without Prior Arrest Group Participant relationships Control relationships Type of offense and participation Drug arrests Participant relationships Control relationships n 29 13 172 24 Ma 5.93 1.85 SD 8.35 3.18 t(39) 2.29b %c c2(3) 20.31d 33.7 0.0 3.39e Z

Note: Comparisons for property, person, and other offenses were not significant. a Indicates mean number of arrests of n. b For comparison with the mean in the row below. c Indicates the percentage of drug arrests of the n total number of arrests. d For comparison of the percentages. e For comparison with the percentage in the row below.

62 the potential influence of adult relationships for adult arrest rates, especially for male friendships and marriageswas also supported, and it contributed to an understanding of the higher arrest rate for participant females. The question is how the effects predicted by the first two hypotheses persisted into the adult years. These predictions came from the results of the Buddy System, which interpreted the beneficial and iatrogenic results as being likely due to friendships formed among participants with and without an arrest prior to referral (ODonnell et al., 1979), an interpretation that has been supported since by many other studies (e.g., Boxer et al., 2005; Karcher et al., 2010; Valente et al., 2007). Based on these studies, it is quite possible that participation in the Buddy System may have altered the peer network of at least some of the adolescents through contact with lower-risk youth and thereby introduced them to a broader range of noncriminal activities, facilitating the beneficial effect found in the results of the first hypothesis. These activities and relationships may have continued into their early adult lives and reduced their chance of arrest. With greater maturity and adult responsibilities, they became less likely to be arrested. In other words, for some of these participants, success within the program altered their activities sufficiently to affect the trajectory of their early adult life, especially their relationships, and that success supported their continued success as adults. A direct test of an altered trajectory of their adult lives is not possible in this study because of the limited information on these relationships in civil and criminal court records. Also, although these court records were essential for tracing adult arrests, they could not provide information on successful, positive relationships and activities. Although a direct test of an altered trajectory interpretation is not possible here, this interpretation is consistent with activity theory. Developed from the work of Vygotsky (1981) on the origination of cognitive development in activities with others, activity theory has been widely used in developmental psychology (Rogoff, 1982), education (Tharp & Gallimore, 1988), cultural psychology (Cole, 1996), and community psychology (ODonnell & Tharp, 1990, 2012; ODonnell, Tharp, & Wilson, 1993). Relationships are formed in settings with people who together engage in the activities of the setting. These shared experiences are intersubjective and thereby affect the cognitions, emotions, and behaviors of the people in each activity setting. This intersubjectivity is a psychological commonality of shared experiences and meaning that creates the means for influence to occur among the participants. This intersubjectivity and mutual influence can be relatively stable over time, if the people and activities remain stable, while change is more likely if activities and people are more diverse or also change (cf. Tharp, 2012). As adolescents become mature adults, people and activities change through employment, romantic relationships, marriage, and adult responsibilities. An altered trajectory interpretation suggests that participation in the Buddy System may have facilitated changes in activities, relationships, and mutual

ODonnell, Williams influence that benefited participants with a prior arrest and contributed to a trajectory that continued into their adult lives. Activity theory explains how such a trajectory may occur; furthermore, if the Buddy System did indeed alter peer friendships, it would have also predicted the results of the first hypothesis. The analyses of the iatrogenic results of the second hypothesis provide additional support for the influence of adult relationships and an altered trajectory interpretation. The arrest rate of female participants without an arrest prior to referral was more than double the rate of the corresponding controls for those with adult relationships listed on their court record. The average number of arrests of their relationships was also significantly higher, and about a third of those arrests were drug related, compared to none for the control relationships. These differences help to explain the higher arrest rate for these females and the absence of a difference for the corresponding males. The high proportion of these drug arrests for offenses committed together suggests a shared participation in drug activities. In activity theory, the shared experiences for criminal activities are just as intersubjective as prosocial activities; they also affect the cognitions, emotions, and behaviors of those participating together in these activities. The mutual influence developed in these shared activities may have occurred in the friendships that some of these females formed with arrested participants in the Buddy System and so facilitated a trajectory of relationships that supported criminal behavior. The potential influence of adult relationships is also apparent from the effects of relationships on the arrest rates of all of the participants and controls combined. These arrest rates were significantly higher, with about three in four people arrested if those in their relationships had been arrested and with about one-third of their arrests for offenses committed with people in their relationships. With the exception of some of the arrests for offenses committed together, the names of the people in their relationships came from the civil court records. This means that the high association of arrest rates with relationships was not an artifact of relationship names found in the criminal court records. Instead, the higher rate and proportion of arrests for offenses committed together illustrate the strong association of relationships and criminal behavior. The type of these relationships was also important. Those with male friendships, including friends with and without arrests, had an exceptionally high arrest rate (90.6%) and the highest rate of offenses committed together (58.6%). In contrast, the corresponding rates were significantly lower for marriage relationships (44.4% and 10.9%, respectively). Overall, the arrest rates were 5 times higher for men with relationships that included arrests, and the lowest arrest rate for men was for those with only women listed among their relationships (24.8%). In addition, of the men without arrests, 83.8% of the people in their relationships were also arrest-free, supporting the potential positive influence of

The Buddy System relationships. These findings support the third hypothesis, on the potential influence of adult relationships for understanding arrest rates. Our datawith almost half of the youth arrested for criminal offenses as adults and about two-thirds of the arrests for serious offensessupport the link of high-risk youth to adult crime. With the effects of the youth mentoring program, positive and negative, continuing well into the adult years, these data show the importance of a focus on adolescent and adult relationships. The higher adult arrest rate for the control group illustrates a potential outcome when there is no intervention. Intervention is especially important for adolescents arrested by their early teens, who are more likely to persist in criminal behavior (Howell, 2008). The lower adult arrest rates of the Buddy System participants indicate that intervention with these young adolescents can be helpful. The form of effective intervention may vary with gender and the risk level of adolescents. Activities between high- and lower-risk adolescents leading to friendships may be particularly harmful for lower-risk adolescent females. The implication for mentoring programs, especially those with a group component, is to design the program to reduce and, if possible, avoid contact between high- and lower-risk adolescents. Doing so reduces the possibility of an iatrogenic effect. The difficulty is that reduced contact with lower-risk adolescents may also reduce the benefit for higher-risk youth. Moreover, when families are dysfunctional; when schools facilitate activities among high-risk youth by policies of tracking, detention, and suspension; and when neighborhoods offer more opportunities for drug use, gangs, and criminal activities than for prosocial events, reducing contact among higher-risk adolescents may not be a viable option. In these situations, it may be more effective to focus on transforming the activities of higher-risk adolescents by providing adult supervision through mentoring, after-school, athletic, and supportive family programs (Smith, Dumas, & Prinz, 2006). These programs can expand youth networks beyond peers by increasing contact and supervised activities with responsible adults. Mentoring programs can affect and be affected by social networks that include adults (Keller & Blakeslee, in press). When successful, the programs may increase the prosocial activities of higher-risk youth. Activity theory postulates that the cognitions, emotions, and behaviors of these youth, as with everyone, are developed in activities engaged with others. If higher-risk youth engage in more prosocial activities with others, they would be more likely to engage in prosocial activities when in contact with lower-risk peers, thereby reducing the possibility of an iatrogenic effect. From an activity theory perspective, the influence of these relationships is dependent on the type of activities in which the youth participate. If high-risk youth were engaged in more prosocial activities with lower-risk peers, there would be fewer opportunities for antisocial influence to occur. Activities are imbedded in our larger social systems. An examination of dysfunctional communities may suggest other

63 possibilities for change. Tharp (2012), in his new delta theory on psychosocial systems, notes the increasing separation of adolescents and adults into different activities in our schools, neighborhoods, and communities. By creating these segregated settings, the influence of adults on adolescents is reduced and frequently replaced by increasing peer influence. If we are to facilitate responsible adult supervision of adolescents, he suggests that we design activities with the adults and adolescents as coparticipants. We offer two recommendations for future programs and research. For programs with adolescents, the critical question is how to design an effective program for high-risk adolescents that does not have an iatrogenic effect. To address this question, we suggest that information on activities and relationships be obtained and used to assess program effectiveness and possible program changes. In the Buddy System, the importance of activities among participants and the peer relationships formed during these activities was not appreciated until after the program, when the results on arrests were analyzed. During the program, the focus was on the relationships between the buddies and youth and the contingencies with problem behaviors. Therefore, specific data on activities and relationships among participants were not obtained. We knew that all of the participants had many opportunities to meet one another and that they often became friends. Subsequently, we found that the connections included those with and without arrests prior to referral. This knowledge led to our interpretation of the arrest data, but the interpretation would be more compelling if we had specific information on participant interactions, not only with their buddies, but also in their neighborhoods and schools. If such data were obtained while the program was ongoing, the data could have been used to inform program changes and increase effectiveness. We recommend that other programs learn about the friendship networks that develop within and outside the program, the activities that the participants engage in together, and the similarities of any changes in emotions, cognitions, and behaviors that the friends experience. These data would not only be important for program effectiveness but could also contribute to knowledge and theoretical development. In this study, online databases were used to successfully obtain arrest and relationship data to follow-up 86.1% of our participants and controls 35 years later. The data showed important associations between arrests and relationships, and with about a third of the arrests for offenses committed together, it is virtually certain that for many, their relationships influenced activities related to their arrests. However, information on relationships based on court records is incomplete. Activities engaged in together, degrees of influence, and much of the qualitative information on relationships are missing. Also missing are relationships without court records. What is needed next are data on these activities and relationships. Successful employment, happy marriages, and achievements are not available in court records to balance the documented troubles in the participants lives.

64 For follow-up research on adult arrests and related measures, the critical question is how participant lives are affected not only by the program but also by the relationships and activities that occur after the program. Data on these relationships and activities could address this question and overcome the limitations of relying on court records. What are their joint activities in their relationships? How did the activities lead to behaviors that resulted in their arrest? How are these activities different for drug, person, and property offenses? What are the gender differences in these activities and relationships? How did these relationships affect their thoughts, emotions, and behaviors? What are the differences in influence among their relationships? What do they recall from their program participation? How did their activities and relationships change from adolescence? What are the trajectories from adolescence that led to or prevented arrest? How do these data differ for those not arrested? What are the activities and relationships that make arrest more or less likely? How do positive activities and other relationships affect these data? For ongoing programs, social media could be used to provide a means for their participants to discuss their lives, especially as related to the program, during and after participation. Rather than just relying on currently available Web sites such as Facebook, programs could construct their own social media sites, designed for their purposes, which would also allow them to control access and ensure security. Program participants could provide ideas and qualitative data while they are in the program and after they leave. Program-designed sites could serve as a resource for participantpractitioner researcher partnerships, information for ongoing program changes, and follow-up data. Similar Web sites could be designed for completed programs. Class alumni Web sites are an example of successful participation for people who identify with their schools and classmates. Similar success is possible for other programs, but it might also be necessary to provide an incentive to ensure a more representative sample of participants. Social media and online databases are making studies such as the ones that we recommend increasingly possible, prospectively and retrospectively. Whether social media or traditional methods are used, it would be important to include samples of those arrested and those not arrested, the people listed in their civil and criminal court records, and their relationships with people not listed in their court records. Differences among these samples would be valuable data to assess the influences on arrests and the prevention of arrest and to make a significant contribution to the research literature. Acknowledgment
Appreciation is extended to Roland G. Tharp for his valuable suggestions, comments, and encouragement.

ODonnell, Williams References


Akers, R. L. (1998). Social learning and social structure: A general theory of crime and deviance. Boston: Northeastern University Press. Arnold, M. E., & Hughes, J. N. (1999). First do no harm: Adverse effects of grouping deviant youth for skills training. Journal of School Psychology, 37, 99115. doi:10.1016/S0022-4405(98)00027-2 Big Brothers Big Sisters. (2011). Starting something since 1904. Retrieved from http://www.bbbs.org/site/c.9iILI3NGKhK6F/b .5960955/k.E56C/Starting_something_since_1904.htm Blokland, A. A. J., & Nieuwbeerta, P. (2005). The effects of life circumstances on longitudinal trajectories of offending. Criminology, 43, 12031240. doi:10.1111/j.1745-9125.2005.00037.x Boxer, P., Guerra, N. G., Huesmann, L. R., & 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, 325338. doi:10.1007/s10802-005-3568-2 Brendgen, M., Vitaro, F., & Bukowski, W. M. (2000). Stability and variability of adolescents affiliation with delinquent friends: Predictors and consequences. Social Development, 9, 205225. doi:10.1111/1467-9507.00120 Butler, S. (1944). The odyssey of Homer. Roslyn, NY: Classics Club / Walter J. Black. (Originally published 1900) Capaldi, D. M., Dishion, T. J., Stoolmiller, M., & Yoerger, K. (2001). Aggression toward female partners by at-risk young men: The contribution of male adolescent friendships. Developmental Psychology, 37, 6173. doi:10.1037/0012-1649.37.1.61 Cole, M. (1996). Cultural psychology: A once and future discipline. Cambridge, MA: Harvard University Press. Dijkstra, J. K., Lindenberg, S., Veenstra, R., Steglich, C., Isaacs, J., Card, N. A., & Hodges, E. V. E. (2010). Influence and selection processes in weapon carrying in adolescence: The role of status, aggression, and vulnerability. Criminology, 48, 187220. doi:10.1111/j.1745-9125.2010.00183.x Dishion, T. J., McCord, J., & Poulin, F. (1999). When interventions harm: Peer groups and problem behavior. American Psychologist, 54, 755764. doi:10.1037/0003066X.54.9.755 Dishion, T. J., & Tipsord, J. M. (2011). Peer contagion in child and adolescent social and emotional development. Annual Review of Psychology, 62, 189214. doi:10.1146/annurev.psych.093008 .100412 Dishion, T. J., Vronneau, M. H., & Myers, M. W. (2010). Cascading peer dynamics underlying the progression from problem behavior to violence in early to late adolescence. Development and Psychopathology, 22, 603619. doi:10.1017/S0954579410000313 Dodge, K. A., Dishion, T. J., & Lansford, J. E. (Eds.). (2006). Deviant peer influences in programs for youth: Problems and solutions. New York: Guilford Press. Dodge, K. A., Greenberg, M. T., & Malone, P. S. (2008). Testing an idealized dynamic cascade model of the development of serious violence in adolescence. Child Development, 79, 19071927. doi:10.1111/j.1467-8624.2008.01233.x Dodge, K. A., & Sherrill, M. R. (2006). Deviant peer group effects in youth mental health interventions. In K. A. Dodge, T. J. Dishion,

Declaration of Conflicting Interests


The authors declared that they had no conflicts of interest with respect to their authorship or the publication of this article.

The Buddy System


& J. E. Lansford (Eds.), Deviant peer influences in programs for youth: Problems and solutions (pp. 97121). New York: Guilford Press. DuBois, D. L., Portillo, N., Rhodes, J. E., Silverthorn, N., & Valentine, J. C. (2011). How effective are mentoring programs for youth? A systematic assessment of the evidence. Psychological Science in the Public Interest, 12, 135. doi:10.1177/1529100611414806 Fo, W. S., & ODonnell, C. R. (1974). The Buddy System: Relationship and contingency conditions in a community intervention program for youth with nonprofessionals as behavior change agents. Journal of Consulting & Clinical Psychology, 42, 163 169. doi:10.1037/h0036260 Fo, W. S. O., & ODonnell, C. R. (1975). The Buddy System: Effect of community intervention on delinquent offences. Behavior Therapy, 6, 522524. doi:10.1016/S00057894(75)80008-6 Gatti, U., Tremblay, R. E., & Vitaro, F. (2009). Iatrogenic effect of juvenile justice. Journal of Child Psychology and Psychiatry, 50, 991998. doi:10.1111/j.14697610.2008.02057.x Glueck, S., & Glueck, E. (1950). Unraveling juvenile delinquency. New York: Commonwealth Fund. Greenwood, P. (2006). Promising solutions in juvenile justice. In K. A. Dodge, T. J. Dishion, & J. E. Lansford (Eds.), Deviant peer influences in programs for youth: Problems and solutions (pp. 278295). New York: Guilford Press. Howell, J. C. (2008). Preventing and reducing juvenile delinquency: A comprehensive framework (2nd ed.). Thousand Oaks, CA: Sage. Karcher, M. J., Davidson, A. J., Rhodes, J. E., & Herrera, C. (2010). Pygmalion in the program: The role of teenage peer mentors attitudes in shaping their mentees outcomes. Applied Developmental Science, 14, 212227. doi:10.1080/10888691.2010.516188 Keller, T. E., & Blakeslee, J. E. (in press). Social networks and mentoring. In D. L. DuBois & M. J. Karcher (Eds.), Handbook of youth mentoring (2nd ed.). Thousand Oaks, CA: Sage. Laub, J. H., & Sampson, R. J. (2003). Shared beginnings, divergent lives: Delinquent boys to age 70. Cambridge, MA: Harvard University Press. Lipsey, M. W. (2006). The effects of community-based group treatment for delinquency: A meta-analytic search for cross-study generalizations. In K. A. Dodge, T. J. Dishion, & J. E. Lansford (Eds.), Deviant peer influences in programs for youth: Problems and solutions (pp. 162184). New York: Guilford Press. McCord, J. (1978). A thirty-year follow-up of treatment effects. American Psychologist, 33, 284289. doi:10.1037/0003-066X .33.3.284 McCord, J. (2003). Cures that harm: Unanticipated outcomes of crime prevention programs. The Annuals of the American Academy of Political and Social Science, 587, 1630. doi:10.1177/0002716202250781 ODonnell, C. R. (1992). The interplay of theory and practice in delinquency prevention: From behavior modification to activity settings. In J. McCord & R. Tremblay (Eds.), Preventing antisocial behavior: Interventions from birth through adolescence (pp. 209232). New York: Guilford Press.

65
ODonnell, C. R. (2005). Juvenile delinquency: Peers, mentors, and activity settings. In C. R. ODonnell & L. A. Yamauchi (Eds.), Culture and context in human behavior change: Theory, research, and applications (pp. 85100). New York: Lang. ODonnell, C. R., Lydgate, T., & Fo, W. S. (1979). The Buddy System: Review and follow-up. Child Behavior Therapy, 1, 161169. ODonnell, C. R., Manos, M. J., & Chesney-Lind, M. (1987). Diversion and neighborhood delinquency programs in open settings: A social network interpretation. In E. K. Morris & C. J. Braukman (Eds.), Behavioral approaches to crime and delinquency: Application, research and theory (pp. 251269). New York: Plenum Press. ODonnell, C. R., & Tharp, R. G. (1990). Community intervention guided by theoretical developments. In A. S. Bellack, M. Hersen, & A. E. Kazdin (Eds.), International handbook of behavior modification and therapy (2nd ed., pp. 251266). New York: Plenum Press. ODonnell, C. R., & Tharp, R. G. (2012). Integrating cultural community psychology: Activity settings and the shared meanings of intersubjectivity. American Journal of Community Psychology, 49 (12), 2230. doi:10.1007/s10464-011-9434-1 ODonnell, C. R., Tharp, R. G., & Wilson, K. (1993). Activity settings as the unit of analysis: A theoretical basis for community intervention and development. American Journal of Community Psychology, 21, 501520. doi:10.1007/BF00942157 Okamura, J. W. (1994). Why there are no Asian-Americans in Hawaii: The continuing significance of local identity. Social Process in Hawaii, 35, 161178. Petras, H., Nieuwbeerta, P., & Piquero, A. R. (2010). Participation and frequency during criminal careers across the life span. Criminology, 48, 607637. doi:10.1111/j,1745-9125.2010.00197.x Reinecke, J. E. (1969). Language and dialect in Hawaii: A sociolinguistic history to 1935. Honolulu: University of Hawaii Press. Rhodes, J. E. (2008). Improving youth mentoring interventions through research-based practice. American Journal of Community Psychology, 41, 3542. doi:10.1007/s10464-007-9153-9 Rhodes, J. E., & DuBois, D. L. (2008). Mentoring relationships and programs for youth. Current Directions in Psychological Science, 17, 254258. doi:10.1111/j.1467-8721.2008.00585.x Robins, L. N. (1966). Deviant children grown up. Baltimore: Williams & Wilkins. Rogoff, B. (1982). Integrating context and cognitive development. In M. E. Lamb & A. L. Brown (Eds.), Advances in developmental psychology (Vol. 2, pp. 125170). Hillsdale, NJ: Erlbaum. Shapiro, C. J., Smith, B. H., Malone, P. S., & Collaro, A. L. (2010). Natural experiment in deviant peer exposure and youth recidivism. Journal of Clinical Child & Adolescent Psychology, 39, 242251. doi:10.1080/15374410903532635 Simons, R. L., Stewart, E., Gordon, L. C., Conger, R. D., & Elder, G. H. (2002). A test of life-course explanations for stability and change in antisocial behavior from adolescence to young adulthood. Criminology, 40, 401434. doi:10.1111/j.17459125.2002 .tb00961.x Smith, E. P., Dumas, J. E., & Prinz, R. (2006). Prevention approaches to improve child and adolescent behavior and reduce deviant peer

66
influence. In K. A. Dodge, T. J. Dishion, & J. E. Lansford (Eds.), Deviant peer influences in programs for youth: Problems and solutions (pp. 296311). New York: Guilford Press. Stannard, D. E. (2005). Honor killing: How the infamous Massie Affair transformed Hawaii. New York: Penguin. Tharp, R. G. (2012). Delta theory and psychosocial systems: The practice of influence and change. New York: Cambridge University Press. Tharp, R. G., & Gallimore, R. (1988). Rousing minds to life: Teaching and learning in social context. New York: Cambridge University Press. Tharp, R. G., & Wetzel, B. J. (1969). Behavior modification in the natural environment. New York: Academic Press. Thornberry, T. P. (1998). Membership in youth gangs and involvement in serious and violent offending. In R. Loeber & D. P. Farrington (Eds.), Serious and violent juvenile offenders: Risk factors and successful interventions (pp. 147166). Thousand Oaks, CA: Sage. Tolan, P., Henry, D., Schoeny, M., & Bass, A. (2008). Mentoring interventions to affect juvenile delinquency and associated problems. Campbell Systematic Reviews, 16, 1112. doi:10.4073/ csr.2008.1610.4073/csr.2008.16

ODonnell, Williams
Valente, T. W., Ritt-Olsen, A., Stacy, A., Unger, J. B., Okamoto, J., & Sussman, S. (2007). Peer acceleration: Effects of a social network tailored substance abuse prevention program among high-risk adolescents. Addiction, 102, 18041815. doi:10.1111/ j.13600443.2007.01992.x Vitaro, F., Brendgen, M., & Tremblay, R. E. (2001). Preventive intervention: Assessing its effects on the trajectories of delinquency and testing for mediational processes. Applied Developmental Science, 5, 201213. doi:10.1207/S1532480XADS0504_02 Vygotsky, L. S. (1981). The genesis of higher mental functions. In J. V. Wertsch (Ed.), The concept of activity in Soviet psychology (pp. 144188). Armank, NY: Sharpe. Warr, M. (1998). Life-course transitions and desistance from crime. Criminology, 36, 183216. doi:10.1111/j.1745-9125.1998 .tb01246.x Wheeler, M. E., Keller, T. E., & DuBois, D. L. (2010). Review of three recent randomized trials of school-based mentoring: Making sense of mixed findings. Social Policy Report, 24, 121. Woodward, L. J., Fergusson, D. M., & Horwood, L. J. (2002). Deviant partner involvement and offending risk in early adulthood. Journal of Child Psychology and Psychiatry, 43, 177190. doi:10.1111/1469-7610.0011

You might also like