This paper reports on the patterns of cocaine use among subjects (N = 427) admitted to a methadone maintenance treatment demonstration project designed to reduce risk for HIV. Assessments were conducted at intake and at approximately 18-24 months after treatment admission. Self-reported data on cocaine use was compared with results of urinalysis tests at both intake and follow-up; 29 subjects who falsely reported no use were recorded as users. Over one-third used some form of cocaine at both intake and follow-up, while approximately 30% abstained at both points. Approximately 20% ceased as well as initiated cocaine use between intake and follow-up. Use of powder cocaine, either alone or combined with heroin in "speedballs," decreased at follow-up, whereas crack use increased. Discriminant function analyses were performed to determine the predictors of the different patterns of cocaine use by type. Receipt of enhanced methadone treatment compared with standard methadone treatment, treatment duration, or average duration of counselor contact appeared unrelated to cocaine use. Cocaine use at follow-up was associated with polydrug and alcohol use, illegal activity, a negative emotional state, and sex work. Crack users were more likely to be African American than nonusers; continuous users of powder cocaine were more likely to also be using heroin than were nonusers; and continuous speedball users were more likely to be women sex workers with high levels of depression. This analysis demonstrated that cessation or continuation of cocaine use after entry into methadone maintenance treatment is not uniform across different types of cocaine.