摘要：Mentoring is one of the most commonly used interventions to prevent, divert, and remediate youth engaged in, or thought to be at risk for delinquent behavior, school failure, aggression, or other antisocial behavior. In this update we report on a meta‐analytic review of selective and indicated mentoring interventions that have been evaluated for their effects on delinquency outcomes for youth (e.g., arrest or conviction as a delinquent, self‐reported involvement) and key associated outcomes (aggression, drug use, academic functioning). Of 164 identified studies published between 1970 and 2011, 46 met criteria for inclusion. Mean effects sizes were significant and positive for delinquency and academic functioning with trends (marginal significance level) for aggression and drug use. Effect sizes were modest by Cohen's differentiation. However, there was heterogeneity in effect sizes across studies for each outcome. The obtained patterns of effects suggest mentoring may be valuable for those at‐risk or already involved in delinquency and for associated outcomes. Comparison of study design (randomised control trial (RCT) or quasi‐experimental (QE)) did not show significant differences in effects. Moderator analysis showed larger effects when professional development was the motivation of the mentors for involvement, but not for basis of inclusion of participants (environmental vs. person basis of risk), presence of other interventions, or assessment of quality of fidelity. We also undertook the first systematic evaluation of key processes that seem to define how mentoring may aid youth (e.g. identification/modeling, teaching, emotional support, advocacy) to see if these related to effects. Based on studies we could code for the presence or absence of each as part of the program effort, analyses found stronger effects when emotional support and advocacy were emphasized. These results suggest mentoring is as effective for high‐risk youth in relation to delinquency as many other preventive and treatment approaches and that emphasis on some theorized key processes may be more valuable than others. However, the collected set of studies is less informative than expected with quite limited specification about what comprised the mentoring program and implementation features. The juxtaposition of popular interest in mentoring and empirical evidence of benefits with the limited reporting of important features of the interventions is seen highlights the importance of more careful and extensive evaluations. Including features to understand testing of selection basis, program organization and features, implementation variations, and theorized processes for effects will greatly improve understanding of this intervention. All are essential to guide effective practice of this popular and very promising approach.
