Past studies of the mediating effects of substance use on the criminal justice involvement of the mentally ill have tended to focus on a single disorder, schizophrenia, and on violent crimes. This study examined the generality of the relationships among psychiatric disorders, substance use, and arrests for violent, nonviolent, and drug-related offenses using data collected for the 2001 and 2002 National Survey on Drug Use and Health. Logistic regression models showed that for violent offenses, the statistical association between serious mental illness (SMI) and arrest across psychiatric diagnoses was substantially but only partially mediated by substance use. For nonviolent offenses and for drug-related offenses, the relationship between SMI and arrest was almost completely mediated by substance use and reduced to statistical nonsignificance. These findings suggest that co-occurring substance use increases the chances a person with any SMI, not just schizophrenia, will be arrested for any offense, not just violent offenses, but that the magnitude of this relationship varies by offense type and, to a lesser extent, by disorder. (Contains 5 tables.)