3. What determines whether early intervention will have a long-term impact? If we assume, for the moment, the validity of the positive effects, a logical next question is what caused them. Juxtaposing the absence of long-term effects in the high-risk study sample with the multiple long-term effects found in the clinically referred sample suggests several hypotheses, all warranting further study. At first blush, the differing findings for the two samples might seem to suggest that early intervention is more likely to produce long-term effects in children with more severe antisocial traits than in those with less severe antisocial traits, a pattern consistent with at least some previous findings (5, 10, 11). However, supplemental analyses by Scott et al. suggest that initial severity does not explain their findings, nor do between-study differences in ethnicity, maternal education, or housing type (public or other). Discerning which factors did make a difference may be difficult, in part because the two study follow-ups differed in numerous ways. For example, in the severe and referred sample, follow-up extended to age 17, potentially permitting more adolescent variability in outcomes (and thus more opportunity to detect treatment versus control differences) than in the follow-up with the high-risk sample, which topped out at age 13. The authors suggest that differences between the long-term findings of the two studies may reflect 1) larger initial intervention effects for the clinic-referred sample, which were thus more likely to be sustained over time; 2) higher motivation by parents of children for whom treatment was being sought; and 3) regression to the mean on antisocial behavior by the high-risk community control group. These three differences, common in treatment-versus-prevention study comparisons, fit the hypothesis that treatment trials with referred children are more likely to produce lasting effects than prevention trials with high-risk children who are not clinically referred. Verifying this or other hypotheses in future studies could help policy makers determine which early intervention investments might be most likely to maximize long-term societal benefit.