Study 1
Bosworth and colleagues (2000) collected data from a middle school near a large Midwestern U.S. city in 1995. The authors described the student population as “diverse,” with 12 percent from the inner city and 20 percent living in the nearby low-income rural area. The middle school randomly placed students on one of three teams at the beginning of each year to organize their academic activities. Each grade (6th, 7th, and 8th) had three teams of 100 to 120 students. Students had little interaction with other teams throughout the school day. The intervention randomly assigned two teams from each grade to the SMART Talk condition and one team to the control condition. The lack of contact between teams, which minimized potential for contamination between treatment conditions, was one reason why researchers conducted randomization at the team level rather than the individual level.
The sample consisted of 558 students. Fifty-four percent were girls; 46 percent were boys. The group consisted of 42, 31, and 27 percent of 6th, 7th, and 8th graders, respectively. The sample was 84 percent white, 9 percent African American, 4 percent biracial, and 4 percent of other ethnicities. The sample’s family arrangements varied: 49 percent of the children lived with both parents; 20 percent lived with a parent and stepparent; 28 percent lived with a single parent; and 3 percent reported other living arrangements (foster care, living with grandparents, etc.).
The outcomes of interest were aggressive behavior, self-awareness, beliefs supportive of violence, self-efficacy, and intentions to use nonviolent strategies. Aggressive behavior was measured through a self-reported 4-item Aggression Scale and a 3-item Conflict Tactic Scale. Self-awareness, beliefs supportive of violence, self-efficacy, and intentions to use nonviolent strategies were all assessed using self-report scales. The study looked at baseline equivalency and performed repeated measures with multiple analysis of variance, with additional multivariate analysis to assess the effects of one semester of intervention. However, the analysis only featured pre- and posttests with no follow-up. The study authors did not conduct subgroup analyses.