There is a powerful association between antisocial behavior and substance abuse. What is still uncertain is whether the association between the two is causal, so that one disorder leads to the other, or is explained by shared symptoms or shared risk factors, or suggests that the two disorders are not distinct, but are actually variants of the same underlying disorder. Each of these hypotheses is shown to be plausible. The paper considers four criteria for causality: precedence, coherence with existing knowledge, dose-related liability, and understandability of mechanisms. Problems are noted with each of these criteria. Conduct disorder as a cause of substance abuse fulfills these criteria more obviously than does substance abuse as a cause of antisocial behavior, but both have plausibility. A similarity is noted between the tasks of deciding whether one disorder causes another and deciding whether early symptom patterns predict the later course of a single disorder. The dearth of information about effect of the early symptom profile on the later course can be overcome with careful study design. Such studies promise important gains in patient management. The necessary data are outlined and instruments to collect such data are noted to be newly published or in development.