This article argues that evaluation of drug courts will benefit not only fr
om an organizing typology (Goldkamp, 1999a, 2000) that focuses research on
the critical structural elements of the drug court model but also from an u
nderstanding of how drug courts are thought to deliver their impact. In dev
eloping a causal model of drug court impact, the analysis separates assessm
ent of impact into two investigations: whether drug courts "work" and how t
hey work. Data from the ongoing NIJ-supported evaluation of the Portland an
d Las Vegas drug courts are analyzed to answer the comparative question of
whether there is an impact (and of what sort) and then to move consideratio
n of the internal elements of the drug court (the black box of drug court t
reatment) through the development of successive theoretical models. The ill
ustrative analyses guided by these models consider the relative contributio
ns of instrumental drug court treatment functions and defendant risk attrib
utes, which contribute importantly to drug court outcomes. The exploratory
findings differ by site, but show some support for the importance of treatm
ent, sanctions and appearances before the drug court judge-and their intera
ction-in lowering the prospects for subsequent rearrest and increasing like
lihood of graduation.