Three methods of analysis were used to determine the effects of the combination of counseling with fluoxetine (20, 40, or 60 mg) and "active" placebo (diphenhydramine, 12.5 mg) randomly assigned. Forty-five cocaine-only dependent subjects were treated as outpatients with "interpersonal" counseling, medication, and drug use monitoring three times per week for up to 12 weeks. Treatment effects were analyzed: first, by comparing the three original assignments and placebo; second, by comparing the placebo group to fluoxetine subjects with detectable fluoxetine/norfluoxetine blood levels and those with no detectable medication blood level; third, by examining relapse prevention versus use cessation through stratifying the subjects into four groups according to fluoxetine or placebo assignment and initial urine cocaine positivity or negativity. All three analyses showed improvement on some measures over time regardless of group assignment. The 60-mg fluoxetine group showed least effectiveness, the group with detectable blood levels had less cravings, and the fluoxetine subjects who were abstinent at the start of treatment were somewhat less likely to avoid relapse than those on placebo.