In meta-analysis the findings from multiple studies are grouped according to some meaningful clustering strategy, and then the studies within those clusters are statistically combined by using weighting and variance-estimating strategies methods that have been developed for that purpose. The 83 empirical evaluations examined in this study involved the following types of interventions for preventing and reducing school violence: administrative techniques (classroom management, building school capacity, schoolwide norms and rules for behavior, etc.); peer mentoring programs; personal growth (non-specific therapeutic programs); self-control; social skills training; peer mediation; educational and academic programs; multimodal programs; and social skills and peer mediation programs. This study found that school-based programs are effective in preventing and reducing violence and other antisocial behaviors. They do this by reducing the mediating conditions and behaviors they seek to alter. All interventions, with the exception of incapacitation, work by altering the presence of trajectory of proximal outcomes that are believed to influence the occasion of the distal outcome that intervention planners seek to avoid. Various method, subject, and implementation variables influenced estimates of program effectiveness. Some of these influences, such as sample similarity, may potentially be controlled by the researcher. The influence of age and risk status, probably is attributable to the greater amount of variance contained in the data produced by these populations. Antisocial behavior is a relatively rare phenomena; ceiling effects probably attenuate many of these estimates of program effectiveness. Finally, programs that experienced fewer implementation problems and lasted longer showed stronger effects than programs that were shorter and had implementation problems. 6 figures and 12 references