The study found that using the evidence on network position yielded by algorithms that cluster agencies into structurally equivalent sets to predict client recidivism was generally a 'blind alley' that yielded few concrete and informative results. Consequently, the study used the simpler measures of agency centrality in each DSO network. A centrality measure indicates the extent to which an agency is connected by relatively short chains to other agencies in the system. Depending upon the regression specification, centrality effects upon recidivism were significant on occasion, although not particularly strong and often not as important as other agency characteristics, such as linkage to the police and courts. The direction of the effects was striking, however. The hypothesis advanced was that a central network position implies access to resources which would enhance treatment effectiveness. It was found, however, that agency centrality was generally associated with increased recidivism. This led to the conclusion that treatment programs occupying a peripheral position with respect to the juvenile justice system and the police avoid the stigmatization effect by diverting juveniles from the routine channels of treatment and punishment while keeping them away from the surveillance of courts and police. Papers developed in connection with the study are appended. Nine references are provided.