Cocaine dependence has proved difficult to treat, whether occurring alone or in combination with opiate dependence. No medication has been demonstrated to be uniquely effective. Fluoxetine was examined as a candidate in two randomized, double-blind, placebo-controlled trials, one with cocaine-dependent patients (study 1) and the other with patients both cocaine and opiate dependent (study 2). It was selected for known specific action, antidepressant effects, minimum side effects, and data showing reduced cocaine effect and self-administration. Clinic visit frequency requirement, a variable with implications for treatment and cost, was also examined in study 1. A total of 228 patients in study 1 and 21 patients in study 2 completed consent and intake procedures. Patients with serious medical or DSM-III-R diagnoses other than cocaine dependence (study 1) or opiate and cocaine dependence (study 2) were excluded. Study 1 patients were assigned to one of two visit frequency schedules (2 or 5 days/week) and one of three medication doses (0, 20, or 40 mg of fluoxetine/day). Study 2 patients received placebo or 20 mg of fluoxetine and 65 to 80 mg of methadone and attended the clinic 5 days/week. All patients participated in individual therapy sessions. Urine screens were conducted twice weekly. A fluoxetine dose response relationship emerged in study 1 for retention with groups from best to worst being placebo, 20 mg, and 40 mg. Dose effect order was the same for both visit conditions. Cocaine use persisted in all groups. The two visits/week condition was correlated with better retention than the five visits/week condition. A significant interaction emerged between intake urine and visit frequency; patients with benzoylecognine screens at intake used cocaine significantly less in the 5 days/week condition, while exhibiting no reduction in the 2 days/week condition. Patients cocaine positive at intake were better retained with infrequent visits. In study 2, a transient reduction in benzoylecognine-positive drug screens emerged for the fluoxetine group. These complementary studies demonstrate that fluoxetine is ineffective in reducing cocaine use or craving. Study 1 also points to setting conditions modulating treatment outcome.