Data were obtained from patients who participated in a larger research and demonstration project that adapted and evaluated a cognitive-behavioral treatment for stimulant abuse within an urban methadone maintenance program. Findings revealed that cocaine use decreased the longer clients remained in treatment and that the amount of cocaine use depended on day of the week, with Saturday typically having the greatest use and Sunday having the least use. Logistic regression analysis showed that several other factors were related to daily cocaine use: peak cocaine craving, resistance to use cocaine, and several triggers or stimuli to use cocaine. These stimuli included receiving money, being offered cocaine, and seeing cocaine and/or related paraphernalia. Even with these variables controlled, however, day of the week and time in treatment continued to be significant predictors. This suggested that other time-varying variables need to be included in order to fully account for cocaine use variation from day to day and that time in treatment is a robust predictor of reduced cocaine use despite the strong influence of craving, external stimuli, and day of the week. 28 references, 5 tables, and 6 figures