Principal Investigator: Sheppard Kellam
Level of Intervention: Universal
Target Population: Elementary school-aged children
References: Dolan, Kellam, Brown, Werthamer-Larson, Rebok, Mayer, Laudoff, Turkkan, Ford, & Wheeler (1993); Kellam & Rebok (1992); Kellam, Rebok, Ialongo, & Mayer (1994); Kellam, Ling, Merisca, Brown, & Ialongo (1998)
Theory (Risk & Protective Factors Targeted):
The Community Epidemiological Preventive Intervention (CEPI), a combination of Mastery Learning (Dolan, Ford, Newton, & Kellam, 1989) and the Good Behavior Game (Dolan, Turkkan, Werthamer-Larsson, & Kellam, 1989) is based on research that links aggressive behavior - as early as first grade - to adolescent antisocial behavior, delinquency, and substance use, especially when aggression is combined with shy behavior. In addition, research has shown that academic failure is related to aggressive behavior and increased risk of depressive symptoms. CEPI seeks to address the proximal outcomes related to shy and aggressive behavior and academic achievement through the combined efforts of the Good Behavior Game and Mastery Learning.
Description of Intervention:
The intervention is conducted over the course of grades one and two. The Good Behavior Game (GBG) is a classroom team-based program designed to improve childrens social adaptation (i.e. reduce aggression and shy behavior) to the classroom relative to rules and authority. Children are assigned to one of three heterogeneous teams in the classroom. During the GBG period, the teams are penalized points whenever a member engages in verbal disruption, physical disruption, is out of their seat without permission, or is otherwise noncompliant. On the other hand, the program rewards teams of classmates for not exceeding maladaptive behavior standards. GBG is conducted 3 times each week for a 10-minute period. Although the criterion for reward remains constant, the length of time that the game is played increases weekly until it reaches a maximum of 3 hours. The timing of the game and the dispensing of rewards are predictable in the early stages of the intervention, but eventually became more sporadic with the time between behavior and rewards gradually extended.
The Mastery Learning component is designed to improve reading achievement through enrichment of the instructional strategies used by teachers in reading curriculums. The program consists of a group-based approach to reading mastery and a flexible corrective process. Students do not progress to next reading level until 80% of the class has achieved 80-85% of the learning objectives in the unit. Weaknesses of individual students are taken into consideration in the corrective process. Grouping strategies and a variety of correctives contribute to flexibility of the program.
Dolan, et al. (1993) reports a sample of 864 students entering first grade in 1985-86 in 19 Baltimore public elementary schools. The sample was 49% male, 64% African-American and 29% white. The GBG sample consisted of 182 students from 8 classrooms, with the GBG internal control composed of 107 students from 6 classrooms. The ML sample consisted of 207 students from 9 classrooms, with the internal control totaling 156 students from 7 classrooms. The external control group consisted of 212 students from 12 classrooms.
In the 6-year followup analysis reported by Kellam, Rebok, Ialongo, & Mayer (1994) 693 students received the intervention for two consecutive years but only 590 were assessed 6 years later. No information is provided on the comparability of the followup sample to the larger group.
5 diverse (e.g. SES, ethnic makeup of neighborhood) urban areas in the city of Baltimore were chosen to participate. Schools and areas were compared. From 19 schools, the 3-4 most similar schools were identified within each area and then randomly assigned to one of three conditions: 1) GBG, 2) ML, or 3) external control condition with no experimental intervention. Individual first-grade classrooms and individual students entering 1st grade were also randomly assigned to intervention or internal control within intervention schools. This design provided both internal (within-school) controls and external (whole school) controls in order to control for within-school contamination of intervention schools and to measure school-level effects.
Measures used to examine intervention impact included teacher ratings, peer nominations and standardized achievement tests. The Teacher Observation of Classroom Adaptation-Revised (TOCA-R, Kellam, et al., 1975) was used as the teacher rating of each childs adequacy of performance on three core classroom tasks. These included social participation (as a measure of shy behavior), accepting authority (as a measure of aggressive behavior), and concentration and readiness to work (as a measure of inattention or concentration problems). The TOCA-R was administered in the fall and spring of first grade. The measures of shy and aggressive behavior specifically were used in the analysis of program impact.
Peer ratings were collected via the Peer Assessment Inventory, a classroom-administered modification of the Pupil Evaluation Inventory (Pekarik, Prinz, Leibert, Weintraub, & Neale, 1976). The Peer Assessment Inventory consisted of six items used to assess the impact of the GBG (3 ratings of aggressive behavior that work well psychometrically as a single score, and 3 ratings of shy behavior that do not work as well as a single score). Internal consistency was .87 for the aggressive behavior scale and .74 for the shy behavior scale.
At six-year followup, the Diagnostic Interview Schedule for Children (DISC 2.25C, Shaffer, Fisher, Piacentini, Schwab-Stone, & Wicks, 1991) was used to identify conduct disorder. The DISC was administered to 184 children consisting of a randomly selected sample of 27 children and 157 others who had screened positive on a conduct problems checklist based on the DSM-III-R (American Psychological Association, 1987).
ANCOVA analyses, controlling for initial levels and considering boys and girls separately showed the GBG had a significant impact on aggressive behavior for boys and girls. For boys the GBG subjects were rated less aggressive compared to the external control group (p<.05) and for girls the GBG subjects were rated less aggressive compared to the internal control group (p<.05).
On peer nominations of aggressive behavior, compared to the internal control group, GBG boys received significantly less aggressive nominations by peers after the intervention (p<.01). No significant differences between GBG girls and controls were found on peer nominations.
For both boys and girls in GBG, teacher ratings of shy behavior were significantly less than internal controls after the intervention (p<.01 for both genders) and significantly less than external controls for girls (p<.01).
Further analysis revealed a specificity of impact such that Mastery Learning affected achievement, but produced no significant behavioral effects. Likewise, the GBG produced significant behavioral outcomes but no affect on achievement.
Follow-up (6 years post-intervention)
Over the follow-up assessment points, comparisons between those children assessed and those not assessed at each point revealed inconsistent differences on some teacher ratings and achievement scores. No main effect reduction in aggression as a result of the GBG was found. For males with higher levels of aggression at first grade, however, there were increasing and significant effects of the GBG at 6th grade. Thus the effect of the GBG varied as a function of aggression severity.
Strengths & Limitations:
GBG is a school-based intervention that specifically targets the interaction between aggressive-shy behavior and academic failure, and poor proximal and distal outcomes for children exposed to these risks. The intervention focuses on the classroom context and utilizes peer/group dynamics, but does not address the role of the family or the larger school ecology. The sample size is adequate and heterogenous enough not to limit the generalizability. Two of the primary sources of data (teachers and peers) were aware of the treatment condition and in some ways had a stake in the outcome, which may have affected internal validity. Though attrition was low, it is unclear how attrition may have affected outcome analysis. Little of the long-term followup data has been reported to date. No information on fidelity of implementation was provided, and there has been no independent replication of the intervention.