Battistich, Schaps, Watson and Solomon (1996) evaluated the Child Development Project, an intervention that addressed 13 positive youth development constructs, including social, emotional, cognitive, behavioral, and moral competencies, bonding, resiliency, self-efficacy, recognition for positive behavior, positive identity, opportunities for prosocial involvement, prosocial norms, and self-determination.
The Child Development Project is characterized by a specific positive youth development goal, to help schools become caring communities of learners. The intervention was designed to become part of the children's overall school experience. The theoretical foundation of the program was based on social learning theory, learning and motivation theory, prosocial development theory, and bonding and attachment theory. The program curricula were integrated with teacher training and new practices to reduce exposure to "old practices." Parents were targeted to build stronger connections between home and school (bonding). The two primary strategies included bringing parents into community-building projects with their children and teachers, and "homeside" activities (bonding, competence). The core program used a comprehensive approach to instructional practices. These consisted of classroom practices such as cooperative learning (opportunities, bonding, competence, self-efficacy, self-determination), a values-based reading and language arts program (prosocial norms, competence), and developmental discipline techniques (self-determination, competence, opportunities). Another component included classroom and whole school elements such as classroom and school community-building activities (bonding, prosocial norms, opportunities).
The study used a quasi-experimental research design, specifically a cohort sequential method involving two demonstration and two comparison schools in each district, at a total of 24 elementary schools from six school districts around the United States. Demonstration schools were selected based on the likelihood they would implement the program; comparison schools were then matched on characteristics of the student populations, and willingness of the school administration to participate. Beginning with baseline assessments in spring 1992, successive cohorts of upper grade students were assessed annually in the spring of each year using group-administered questionnaires. Two sets of analyses examined program effects on student drug use and delinquency. The first set examined condition differences and the second set examined these results in the context of program implementation levels. In the second set of analyses, the 12 demonstration schools were classified into three groups of four schools each -- high, moderate or low implementation -- based on changes from baseline in six observational measures of program practices. School districts included three large urban districts, one small urban district, one suburban district, and one rural district. The authors described the student populations at these schools as varying greatly. The range of students receiving free or reduced school lunch was from 2% to 95%; the sample's ethnic composition over three years of measurement ranged from 39-54% for Caucasian children, 17-23% for African-Americans, 21-27% for Hispanics, 5-10% for Asians, and 2-3% "other" youth (defined as Native American, Alaskan Native, Pacific Islander, and "other"). Between zero and 32% of participants spoke limited-English or no English, and average achievement levels ranged from the 24th to the 67th percentile on standardized tests. The overall sample (n = 1645) included slightly fewer boys (48%) than girls (52%). The overall program included students from the third through the sixth grades; assessments for this evaluation included the fifth and sixth grade cohorts, with average age 11 to 12 years old.
Results of the Intervention
Assessments of drug use and delinquent behaviors were limited to students at the top elementary grade (fifth or sixth) at each of the schools. The student questionnaire results showed that alcohol use significantly declined (4%) at the demonstration schools during the first year of the intervention and stayed the same during the second year (F(2, 4510) = 4.19, p <.02). In contrast, alcohol use increased significantly at the comparison schools during the same period (2-3% increases each year; t(4715) = 2.49, p<.02, ES= .12). Changes in the prevalence of marijuana use and vehicle theft showed similar trends but were not statistically reliable. Tobacco use declined among students at both demonstration and comparison schools over the course of study (F(2,4524) = 11.52, p<.001). The authors also presented results of an extensive implementation analysis of the 12 demonstration schools, which were classified into low/medium/high implementation based on changes from baseline in six observational measures of program practices: teacher warmth and supportiveness, emphasis on prosocial values, encouragement of cooperation, elicitation of student thinking and expression of ideas, use of extrinsic control, and student autonomy and influence. Differences in youth outcomes associated with implementation level were discussed at length, and conclusions about effectiveness were linked with the high implementation schools. However, most of these could not be determined to be statistically significant due to, among other things, reduced sample sizes. There were, however, significant decreases for program students in high implementation schools compared with controls in marijuana use (F(2, 1368) = 8.56, p <.001), carrying weapons (F(2, 1375) = 7.28, p <.002), and vehicle theft (F(2,1368) = 7.11, p <.002). In addition to these outcomes for this large sample, another evaluation (Battistich et al., 1989) of an earlier, smaller (n=521) sample of Child Development Project participants reported that program children were significantly more accepting of, and accepted by, their classmates, were less lonely, and were lower in social anxiety than children in comparison schools.
Greenberg (1998) and The Conduct Problems Prevention Research Group (1997) evaluated the Fast Track Prevention Project, a multi-site, multi-year intervention that addressed 10 positive youth development constructs, including social, emotional, cognitive, and behavioral competence, self-efficacy, resiliency, bonding, recognition for positive behavior, opportunities for prosocial involvement, and prosocial norms.
The comprehensive intervention is guided by developmental theory on youth competence and the multiple influences through which antisocial behavior manifests. Fast Track is based on a unified model of prevention that consists of both universal and selective components, with some of the selective components individualized to the needs of the child and the family. The universal component is the PATHS teacher-led classroom curriculum. The selective intervention components were administered only to the high-risk sample, and included PATHS, parent training groups, home visits, child social skill training groups, child tutoring in reading, and peer pairing in the classroom. The Fast Track curriculum in first grade had 57 lessons of the main curriculum, taught approximately three times a week with each lesson lasting 20-30 minutes. Forty percent of the lessons focus on skills for understanding and communicating emotions, thirty percent focus on skills for increasing positive social behavior; and thirty percent focus on skills for self-control and social problem-solving. Parent and child training groups were weekly in grade one, biweekly in grade two, and monthly thereafter for the rest of the project. Teacher training included a two and a half day workshop and weekly consultation and observation with project staff.
Approximately 13 schools within each of four sites (Durham, North Carolina; Nashville, Tennessee; Seattle, Washington; and rural central Pennsylvania) were selected based on community risk factors and invited to participate in the intervention. Schools were randomized into intervention and control participants (after obtaining building consensus to participate) and the intervention was conducted in three successive years with three cohorts of first graders. There were 201 classrooms and 184 matched comparisons across the three cohorts. For the high-risk sample, a multi-stage screening of all kindergarten children based on teacher and parent ratings yielded the top 10% of high-risk children. They were then contacted for pre-intervention assessments and upon completion of these, included in the total sample with the condition based on the school they entered for the first grade. This resulted in three successive yearly cohorts for a total of 448 intervention and 450 control group children. The high-risk sample was approximately 47% Caucasian, 51% minority students, generally African-American, and two-thirds male. Fifty-five percent of children received free or reduced lunch and the mean reading percentile was in the 49th percentile. No significant differences existed between intervention and control schools on any of these characteristics.
Results of the Intervention
The evaluations reported findings for both the universal and high risk samples. The full study results were reported through teacher, peer, and independent observer ratings. Results for the high risk sample were reported through teacher and parent measures. In the full study, the teacher ratings showed outcomes in favor of the intervention group on measures of accepting authority (F(1,333) = 3.6, p<.04), and a higher mean liking score (F(1,333) = 3.8, p<.05). Peer ratings showed that intervention classrooms had lower peer aggression scores (F(1,377) = 5.42, p=.02), and less hyperactivity (F(1,377) = 4.66, p=.03) than controls. Observer ratings of positive classroom atmosphere were significantly higher for the intervention classrooms (F(1,279) = 5.6, p=.01). Sub-ratings of these scales showed improvements for students' appropriate expression of feelings (F(1,279)=4.3, p<.05), classroom enthusiasm (F(1,279) = 8.2, p<.01), and the classroom's ability to stay focused and on task (F(1,279) = 9.5, p<.005).
Analyses of the high risk sample using growth curve models showed the impact of the intervention on two conduct problem measures provided by parent daily report and teacher observations of classroom adaptation. Outcomes showed that the control group increased in disruptive behavior at school compared with intervention children, while intervention group children showed greater reductions in conduct problems at home compared with controls. Although the intervention and control groups started at the same level of disruptiveness, the control group increased on this dimension while the intervention group stayed the same or declined slightly. Parent data showed similar patterns of greater decline in conduct problems at home for the intervention group youth.
Analyses of nine mediating variables measuring the mediating role of child and parent competencies (six child and three parent competency measures, tested as a set) were conducted using the average of the end of second and third grade assessments. Investigators used structural models linking the intervention condition to each youth behavior measure and found significant results for the mediating variables: improvements in both child and parental competency factors did in fact mediate the decreases in child conduct problems. The results of the growth curve data underscore an important aspect of positive youth development, namely that intervention effects are gradual and cumulative and operate along a developmental (time-based) trajectory.
Results of the Implementation Analyses
The analyses examined the effects of dosage and used four ratings of the quality of implementation: teacher's rated skill in teaching PATHS concepts, skill with which teachers modeled and generalized PATHS concepts throughout the day, the quality of teacher classroom management skills, and teacher's openness to consultation with the educational coordinator. For dosage there was a significant main effect on the sociometric rating of prosocial behavior. The teachers' rated skill in teaching and modeling PATHS concepts, the skill with which teachers modeled and generalized PATHS concepts, and classroom management skills were all significantly related to scores for authority acceptance, cognitive concentration, and social competence. Classroom management skill was also related to the sociometric score on prosociality. Openness to consultation was significantly related to cognitive concentration, social competence, prosociality, and to hyperactivity.
The overall results from this intervention suggest some interesting questions for future studies of positive youth development. As the authors (Conduct Problems Prevention Research Group,1997:13) said: "The interpretation of these findings is a bit complicated by the fact that the universal intervention was delivered along with the selective intervention for children with early behavior problems. Although analysis with and without the high-risk children showed similar patterns, the classroom scores may not be entirely free of the impact that the selective intervention may have on the high-risk children. Improvements in the high-risk children may reduce teacher stress, improve peer relations and thus also possibly affect outcomes."
Eron, Guerra, Henry, Huesmann, Tolan and Van Acker (1997) evaluated the impact of Metropolitan Area Child Study, a multi-year prevention field trial that addressed 10 positive youth development constructs, including social, emotional, cognitive, behavioral and moral competencies, bonding, resiliency, self-efficacy, opportunities for prosocial involvement, and prosocial norms.
This intervention is a combined universal and selective prevention program designed to influence norms affecting aggression in a general population, and to lower aggressive behaviors in a high-risk sample. The intervention goal was to modify the youth's cognitive system while influencing aspects of her social systems (the classroom, the school and the family) shown to be key in the learning of aggressive behaviors. The intervention used different combinations of positive youth development strategies across conditions. In the first condition, the classroom intervention integrated three components including a teacher education program, collaborative support strategies from project staff, and a social-cognitive curriculum called "Yes I Can." The educator strategies included the development of prosocial behavior in students (prosocial norms, recognition), cultural sensitivity (competence, bonding), and proactive behavior management (self-efficacy, competence, resiliency). The curriculum strategies consisted of 40 one-hour lessons taught by teachers over two years, and included teacher manuals and student workbooks. Program themes included self-understanding (self-efficacy, resiliency), relationship of self to others (bonding), and moral beliefs (competence, prosocial norms). The second condition was a classroom-based general enhancement intervention, and an intensive small-group component in which youth received social-cognitive and peer-relationship training in a small group. These strategies targeted children's cognitive attributions, beliefs and scripts, and gave them modeling, role-play and leadership opportunities to practice new scripts in a non-threatening setting. Small groups met during the regular school day once a week for 28 weeks over two years (12 weeks in first year and 16 weeks in the second year). The third condition included the classroom program, the small group intervention, and a family intervention. The family relationship component emphasized parenting management skills, enhanced communication (bonding, competence), and family cohesiveness strategies including family problem-solving, monitoring, reframing and coping skills (competence, bonding, prosocial norms, opportunities).
The intervention used an experimental design. Two urban communities (Chicago and Aurora) with many high-risk children were selected within a larger metropolitan area. Schools were recruited through an extensive screening process, and 16 elementary schools participated (from an original group of 130 that applied). These were randomly assigned to each of four experimental conditions including three intervention levels and a control condition, as described. Four schools received each of the three interventions and there were four control schools that received none. Schools were blocked on the basis of their ethnic composition, geographic location and aggressiveness rating level of the children. The four blocks were Chicago Hispanic, Chicago African American, Aurora Hispanic, and Aurora Mixed. One block of four schools was randomly assigned to each condition. Two schools (one in Chicago, one in Aurora) were replaced because of administrative changes before the first year, and one school was replaced after the third year of the intervention due to administrative issues. The attrition analysis showed that from baseline to posttest, attrition was about 44% (range of 39% in Level C to 50% in Level A), yielding a final sample of 789 children. Initial aggression scores indicated that aggression was significantly higher for the group that dropped-out. The overall sample was studied in three two year waves: 1991-1993, 1992-1994, and 1993-1995. This included an early intervention consisting of four cohorts with second-third or third-fourth grade students, and a late intervention consisting of two cohorts of fifth-sixth grade students. The sample was 40% African-American, 40% Hispanic and 20% Caucasian. Two-thirds of the children qualified for a free lunch program, with significant variance across ethnic groups (African Americans were the mostly likely to qualify, Caucasians were the least). Low socioeconomic status was significantly correlated with aggression levels at baseline. The baseline aggression scores of the study sample were high compared with national averages. There were no significant differences in ethnic composition, SES, levels of aggression, and beliefs about aggression across the four conditions.
Results of the Intervention
Published results for the Metropolitan Area Child Study showed mixed outcomes. Pretest aggression scores were used as a covariate. Children who were already moderately prosocial and received the small group intervention were found to have significantly increased prosocial behavior at posttest (F(3,361) = 4.04, p<.01). Prosocial behavior was found to have a significant moderating effect on posttest aggression scores for the early intervention condition. In the early intervention condition, control group participants who were low in prosocial behavior and in aggression at pretest were found to have high levels of aggression at posttest; this was not true for intervention children, which the authors suggested meant the program prevented the development of high levels of aggression in this sub-group of children. Late intervention (grades 5-6) children who were high in pretest aggression had significantly lower posttest aggression scores than control children (F(3,257) = 4.8, p<.003). However, the late intervention children who scored as moderately aggressive at pretest scored worse on the posttest than comparable control group children.
In addition to these published evaluation data, more recent unpublished evidence showed significant effects for decreased aggression for the full sample, delivered early (second and third grade). Results showed the early intervention improved the program group's on- task behavior and stopped deterioration of prosocial behavior. Both the early and late versions prevented normal deterioration in academic achievement, compared with control youth. These effect sizes ranged from .15 to .33 (Tolan et al.,1998).
Kirby, Barth, Leland and Fetro (1991) evaluated Reducing the Risk, an intervention that addressed nine positive youth development constructs, including social, emotional, cognitive, behavioral and moral competencies, self-efficacy, opportunities for prosocial involvement, prosocial norms, and self-determination.
The program strategies combined cognitive behavioral skills training (competence, self-efficacy), a social inoculation focus (self-efficacy, self-determination), teacher and peer role modeling (opportunities, prosocial norms), and organizational strategies designed to influence students' motivation and peer group variables associated with delaying sexual activities and effective contraception. The curriculum structured parental involvement in children's learning and skills acquisition through multiple homework discussion strategies. The program exposure was 15 class periods and some parent-child discussion of an unspecified amount, with teachers attending a three-day training session. Implementation was assessed through observations and ratings by the evaluation team, and student and teacher perceptions of the curriculum usefulness. The evaluation collected data from pretest to posttest and over an 18-month follow-up period.
The evaluation used a quasi-experimental design with partial random assignment and comparison groups. Forty-six classes of students taking a mandatory health education class were assigned to either treatment or comparison groups. Some of the classes were randomly assigned, and others were not. Approximately half of the program and control classes that were taught by the same teachers; the other half were divided as program classes taught by one group of sexuality educators and comparison classes taught by a different group. Thirteen urban and rural schools in California participated with a total of 1033 students in 23 classes (586 in the intervention group; 447 in the comparison group) surveyed at pretest, and 758 completing both the posttest and follow-up questionnaires. The authors (Kirby et al., 1991: 256) note: "Because most comparison classes were receiving whatever sexuality instruction those teachers were already providing, this evaluation measures the impact of the RTR curriculum relative to other sexuality curricula, not the overall impact of RTR alone." The sample consisted of 53% female youth and 47% male youth, of whom 62% were Caucasian, 20% were Latino, 9% were Asian, 2% were Native American, 2% were African American, and 5% were of unspecified ethnic identity. The attrition analysis showed a 27% rate for both the intervention and comparison groups, and those lost to attrition differed in several ways from the remaining participants. At pretest they were more likely to be older, male, sexually experienced, not live with both parents, have poor grades, not have talked to their parents about abstinence, and failed to use birth control because sex was unplanned.
Short-Term Results of the Intervention
The initial study results were based on posttest outcome measures at six months of contraceptive knowledge, perception of peer sexual prevalence, sex initiation, and unprotected sex. These indicators were assessed by the students, teachers and parents. At posttest measurement, the intervention group had significantly greater contraceptive knowledge than the control group (76% vs 65%, p<.001). There were greater increases among the intervention students in communicating with their parents about abstinence (66% vs 61%, p<.01) and birth control (53% vs 41%, p<.01). The comparison group believed at posttest that more of their peers were having sex than at pretest (3.1% vs 3.3%, p<.01), while there was no change in this belief for the intervention group.
Long-term Results of the Intervention
Results for the long-term follow-up showed sustained behavioral effects for the intervention. Significant results were found in lower rates of initiation of intercourse among intervention participants who had never had intercourse (29% vs 38%, p<.05). Measures of contraceptive practices showed no significant overall differences between the two groups for increases in contraceptive use from pretest to posttest, but there were significant differences in some sub-groups. Only 4% of females and 3% of low-risk teens in the program group had unprotected intercourse, compared with 16% of females and 11% of low-risk youth in the comparison group (p<.001). Among all lower risk youth independent of previous sexual experience, youth in the program group were significantly less likely to have had unprotected sex at last intercourse, and less likely to have had unprotected sex all or most of the time. At pretest, 11% of both intervention and comparison groups had engaged in unprotected sex; at 18 months only 13% of the intervention group were not using contraception, compared to 23% of the comparison group (p<.05). Other continued effects seen at 18 months for the students in the program group included greater levels of contraceptive knowledge (p<.001) and communication with parents about abstinence (p<.05).
Hawkins, Catalano, Kosterman, Abbott, and Hill (1999) evaluated the Seattle Social Development Project, a multi-year interventionthat addressed nine positive youth development constructs including social, emotional, cognitive and behavioral competencies, bonding, self-efficacy, opportunities for prosocial involvement, recognition for positive behavior, and prosocial norms.
The intervention sought to promote academic achievement and other competencies, and to prevent a range of negative youth outcomes including crime, teen pregnancy, drug abuse, school failure and dropout, by to increasing prosocial bonding, which is hypothesized to be a mediating construct of the desired outcomes. Components included teacher training in methods of classroom instruction and management, parent training in developmentally appropriate methods of parenting, and social competence training for youth. In the first and second grades, parents in the full intervention condition were offered the seven-session "Catch 'Em Being Good" curriculum for child behavior management skills. In the spring of second grade and again in third grade, these parents were offered the four session" How to Help Your Child Succeed in School" curriculum. In grades five and six, full and late intervention parents were offered the five-session Preparing for the Drug Free Years curriculum. In grade six, students in the full and late treatment groups received four hours of training from project staff in recognizing and resisting social influences to engage in problem behaviors and generating positive alternatives for staying out of trouble. Teachers in intervention classrooms received five days of inservice training per year on proactive classroom management, interactive teaching, and cooperative learning methods to use in managing and teaching their classes.
The study used a quasi-experimental design with comparison groups. The population overrepresented schools located in high-crime, high-poverty neighborhoods, but youth within classrooms in these schools were treated universally. To assess the effects of full intervention and late intervention, a nonrandomized controlled trial with three conditions was created. The full intervention group received the intervention package from grade one to six. The late intervention group received the intervention package in grades five and six only, and the control group received no special intervention. This design was created in 1985 by nesting an intervention study begun in 1981 at first grade entry within the longitudinal panel study. For the present study, schools were assigned to conditions in the fall of 1985 and from that point, all fifth grade students in each school participated in the same interventions. The full intervention group consisted of all students who were randomly assigned to intervention classrooms in grades one through four in eight elementary schools participating in the earlier experimental study, and who remained in schools assigned to the intervention condition in grades five and/or six in the present study. The late intervention group consisted of students in intervention schools who were in intervention classrooms in grades five and six only, some of whom were controls in the earlier intervention study. The control condition consisted of students in schools assigned to receive no intervention in grades five and six and who were not in intervention classrooms in grades one through four. The final sample included 643 fifth grade students assigned to three conditions whose parents consented to their involvement in the follow-up study, which was conducted six years after intervention at age 18. Of these 643 students, 598 were successfully tracked and interviewed (93%). The demographic and ethnographic composition among three experimental conditions ranged from 47-53% male, 42-46% Caucasian, 21-30% African-American, 18-24% Asian, 3-7% Native American, and 1-5% of other ethnic identity. Participants' eligibility for free lunch averaged 56%. At age 18 the three study groups did not differ significantly on residential stability, socioeconomic status, or proportions who were male, Caucasian, non-Caucasian, or from single-parent households. Implementation quality was assured by teachers in both sets of classrooms being observed for 50 minutes on two different days in the fall and spring of each year, and given scores rating fidelity to intervention practices. Data were also collected using questionnaires for youth, parents and teachers; data on delinquency charges in King County Juvenile Court, California Achievement Test results, grade point averages, and school disciplinary action reports through age 17 from the Seattle School District.
Long-term Results of the Intervention
The findings reported when the students were 18 (six years post intervention) showed positive effects for the full treatment intervention group compared to the control group on many of the school, delinquency and sexual behavior outcomes. Students in the full intervention reported significantly stronger attachment to school (p=.006), improvement in self-reported achievement (p=.01) and significantly less involvement in school misbehavior (p=.02) than controls. While no effects were shown for either the full or late intervention groups for lifetime prevalence of cigarettes, alcohol, marijuana or other illicit drug use at age 18, significantly more subjects in the control group than in the full intervention group had committed violent acts (48.3% vs 59.7%, p=.04), reported heavy alcohol use in the past year (15.4% vs 25.6%, p=.04), had engaged in sexual intercourse (72.1% vs 83%, p=.02), or had multiple sex partners (49.7% vs 61.5%, p=.04).
Analysis of the interactions between poverty and intervention condition showed that the full intervention was significantly more effective for poorer children in positively affecting attachment to school and in reducing having to repeat a grade. The intervention had significantly greater effects for working and middle class youths in reducing the lifetime prevalence of pregnancy or of having or fathering a baby. Most observed intervention effects did not differ by gender; however, the full intervention had significantly greater effects on preventing males from repeating a grade and engaging in sexual activity.
Weissberg and Caplan (1998), and Caplan, Weissberg, Grober, Sivo, Grady and Jacoby (1992) evaluated the Social Competence Program for Young Adolescents, which addressed nine positive youth development constructs including social, emotional, cognitive and behavioral competencies, bonding, prosocial norms, self-efficacy, recognition for positive behavior, and resiliency.
The program's six units were designed to enhance personal and interpersonal effectiveness (social-emotional competence and bonding) and to prevent the development of maladaptive behavior. The intervention did this through teaching and reinforcing students' developmentally appropriate skills and information (cognitive and behavioral competence, self-efficacy, recognition), fostering prosocial and health-enhancing values and beliefs (prosocial norms) and creating environmental supports (all forms of competence, bonding, and resiliency). The 12-week intervention was a 16-session competence promotion program that used highly structured, scripted 45-minute lesson plans in conjunction with applications to promote competence and prevent substance abuse.
Implementation strategies included training for teachers and undergraduate aides in 10 90-minute workshops before program implementation, and continuing throughout the 16-week intervention. They also received on-site consultation and coaching during classroom lessons, and following the 16-week program received three monthly two-hour training to focus on ways to help maintain SCP skills by students. Two observers independently rated the quality of program implementation in each classroom.
The evaluation used a quasi-experimental research design with comparison groups. The program was described to teachers at four middle schools who were then given the opportunity to voluntarily participate. Teachers who volunteered had their classrooms assigned to a control or program condition based on scheduling and comparability of academic-ability grouping levels across conditions. Separate analyses were conducted to compare the outcomes associated with high versus low implementation fidelity. Tests of the groups' equivalence at baseline indicated they were comparable to one another on all but eight (five favoring the intervention group, three favoring the control group) of 28 variables at pretest. The sample consisted initially of 447 fifth to eighth grade students from 11 program (n=238) and nine control (n=183) classes in four urban, multi-ethnic schools. The final sample (n=421) of 210 male and 211 female students from low- and middle-income families included 178 Caucasian students, 167 African-American, 72 Hispanic, and four students of other, unidentified ethnic identity. The authors provided analyses of group differences showing that control and program groups did not differ with respect to grade level, gender or race and that students had comparable academic achievement levels. However, the groups did differ on 8 of 28 variables at pretest. The attrition analysis showed high retention levels across groups, with the study losing only 26 students (5.8% of the sample) whose parents withheld permission.
Teacher observations were made by both primary and secondary teachers (who were hypothesized to be blinded to treatment condition) to help deal with the potential of inherent teacher bias. After the main program impact analyses, a related set of analyses were performed to assess whether the quality of program implementation influenced student outcomes by comparing pre-to-post changes for students from well-implemented program classes, less well-implemented program classes, and the control group. Three assessment instruments yielded 16 measures of students' problem solving skills, conflict resolution attitudes, and self-reported assertiveness in response to hypothetical situations.
Results of the Intervention
The evaluation found significant positive effects on minor delinquency (36.8% vs 2.8%, (F(1,399) = 4.37, p<.05) and increases in or maintenance of positive behaviors including improved behavioral conduct (F(1,399) = 8.57, p<.01), involvement with peers (F = 6.02, p<.05), and social-acceptance ( F = 4.17, p<.05). Significant changes were noted for the intervention group for social problem-solving (F(9,168) = 4.92, p<.001), and improvements in attitudes for conflict resolution strategies (F(6,171) = 2.97, p<.01). Program participants improved more than controls in generating a greater number of cooperative solutions (F(2,192) = 16.14, p<.01), more effective solutions (F(2,192) = 5.52, p<.01) and more planful solutions (F(2,192) = 8.08, p<.001). Program students generated significantly lower percentages of aggressive (F = 4.94, p<.05) and passive (F = 11.29, p<.01) solutions from pre to posttest when compared with control group students. Program students liked peers who resolved conflicts in assertive ways significantly more than control students (F(1,176) = 4.07, p<.05); endorsed cooperative strategies more than control students (F(1,175) = 4.24, p<.05); and were shown to respond more assertively rather than passively or aggressively to hypothetical situations (F(1,311) = 7.06, p<.01). Primary teacher ratings showed that relative to controls, program students improved significantly in behavioral conduct (F = 17.54, p<.001) but not in social acceptance by peers.
Results Based on Implementation Analyses
Independent of implementation quality, program students showed significant gains relative to controls in solution effectiveness, number of planful solutions and percentage of cooperative solutions, and decreased more in the percentage of passive solutions. Students in well-implemented program classes improved significantly more than those in lower-quality implementation classes and control classes on the number of alternative solutions generated to problem situations (F(2,192) = 13.01, p<.001), the extent of endorsing assertive (F(2,193) = 3.58, p<.05) and cooperative (F(2,193) = 4.09, p<.05) conflict-resolution strategies, and on secondary teacher ratings of behavioral conduct (F(2,366) = 8.95, p<.001) and social acceptance (F(2,366) = 15.88, p<.001).
Slavin et al. (1996) evaluated the impact of Success for All, an intervention that addressed eight positive youth development constructs, including social, cognitive, and behavioral competencies, bonding, self-efficacy, opportunities for prosocial involvement, recognition for positive behavior, and prosocial norms.
The program has been described elsewhere as based on reading achievement, but its focus on academic success is grounded in a more comprehensive philosophy of what it takes to promote positive youth development in children. The program authors refer to their concern with a focus on the success of every child, and a philosophy of "neverstreaming," i.e., not separating special needs students from the mainstream. The program formula links academic success and cognitive competence to developing a child's overall capacity for self-respect and self-efficacy, concluding that in their absence, the ingredients necessary for successful positive youth development disappear: "When a child fails to read well in the early grades, he or she begins a downward progression" (Slavin et al., 1996: 42). The program addressed cognitive competence in several ways, through reading performance achievement, and through rehearsing strategies for self-assessment and self-correction. Other positive youth development strategies included one-to-one tutoring, cooperative learning, assessments every eight weeks, vision/hearing screenings, working with parents and social service agencies to ensure attendance, parenting skills workshops, providing medical services and eyeglasses, and helping with behavioral problems. This six-year intervention targeted Grades K-5 in a sample of 110 schools. At the time of evaluation the program was being implemented in 300 schools in 23 states.
The quasi-experimental research design was a 'multi-site replicated experiment' design in which groups were matched rather than randomly assigned, and each grade level cohort was specified as the unit of analysis. For specific site studies the unit of analysis was the individual student; however, the comprehensive evaluation reported here chose the multi-site replicated experiment method of analysis which specified each grade level cohort as the unit of analysis. Each intervention school was matched with a comparison school based upon variables including poverty-level percentage of students qualifying for a free lunch, achievement level, and ethnicity of the student body. Individual children in program schools were matched with children in non-program schools based on either scores on district-administered standardized tests or scores on the Peabody Picture Vocabulary Test. The overall findings for 23 evaluated schools constituted 55 experimental and 55 control cohorts, with each cohort made up of 50-150 students. The sample of almost all African-American students in first grade were 75-96% eligible for a free lunch program.
Results of the Intervention
The results showed significant differences in favor of the intervention group for reading scores at all grade levels. The Success For All program yielded statistically significant positive effects on every outcome measure at every grade level (p<.001), which averaged around half a standard deviation per grade level. Students in the lowest 25% of their grades showed effect sizes averaging one standard deviation. These effect sizes progressively increased with each year of implementation of the program. The intervention also showed significant positive gains for Asian (mostly Cambodian) students learning English as well as Spanish speaking students who were provided with a Spanish version of the Success for All curriculum. The overall comparison of Success for All to Reading Recovery, a similar reading program based upon one-to-one tutoring, showed only slightly better results for Success for All when looking at results for all children. However, an analysis of only special education children showed substantial differences in favor of Success for All students (effect size = .77).