General population survey data are used to disaggregate the associations of substance use disorders with suicide attempts in order to evaluate a number of hypotheses about the processes leading to these associations. Data are from the US National Comorbidity Survey (1990-1992). Discrete-time survival analysis is used to study the effects of retrospectively reported temporally prior substance use, abuse, and dependence in predicting first onset of suicidal behavior. Alcohol and drug use predict subsequent suicide attempts after controlling for sociodemographics and comorbid mental disorders. Previous use is not a significant predictor among current nonusers. Abuse and dependence are significant predictors among users for three of the 10 substances considered (alcohol, inhalants, and heroin). The number of substances used is more important than the types of substances used in predicting suicidal behavior. Disaggregation shows that the effects of use are largely on suicidal ideation and nonplanned attempts among ideators. In comparison, the effects of use on suicide plans and planned attempts among ideators are not significant. Clinicians need to be aware that current substance use, even in the absence of abuse or dependence, is a significant risk factor for unplanned suicide attempts among ideators.