Little is known about gambling rates of drug users recruited from drug treatment compared with those recruited from the community. We use the Diagnostic Interview Schedule (DIS) to provide lifetime prevalence estimates of problem gambling (i.e., at least one gambling problem) and DSM-III-R pathological gambling (i.e., at least four gambling problems) and describe the association between gambling and psychiatric disorders for drug users recruited from drug treatment settings (n = 512) and from the community (n = 478). We also report the relative risk of being a recreational and problem gambler in this sample. The sample was first interviewed in 1989-90 as a part of two NIDA-funded St. Louis-based studies. The prevalence of problem gambling in the overall sample was 22% and the prevalence of pathological gambling was 11%. There were no statistically significant differences in problem and pathological gambling rates for subjects recruited from drug treatment and those recruited from the community. The conditional prevalence rates, that is, the rate of problem and pathological gambling only among gamblers were 27% and 13.5%, respectively. Major findings indicate that problem gambling was associated with Antisocial Personality Disorder (ASPD), even after controlling for recruitment source and socio-demographic characteristics. In fact, when examining the temporal order of these disorders, we found that pathological gambling was always secondary to ASPD, occurring on average 11.4 years after the onset of ASPD. Problem gamblers, compared with everyone else, were more likely to be male, African-American, recruited from drug treatment, have ASPD and be dependent on illicit drugs. Multinomial logistic regression analysis predicted the relative risk of being a recreational and problem gambler (compared with a nongambler) in this sample according to socio-demographics, ASPD, and dependence on illicit drugs. Results imply that screening for gambling problems will need to be broad-based among drug users.