A properly implemented agonist treatment regimen should improve retention and reduce illicit drug use. Cocaine-dependent subjects (N = 128) were enrolled in a 12-week randomized, double-blind, placebo-controlled trial. In the multistage dosing design, subjects initially received placebo (PBO) or 15 to 30 mg of dextroamphetamine sulfate, sustained-release capsules. At week 5, the dose doubled to 30 mg or 60 mg for active groups. Subjects attended the clinic twice a week, provided urine samples, obtained medication, and had one behavioral therapy session a week. Retention was best for the 15- to 30-mg group, whereas the proportion of benzoylecgonine-positive urine screens was, from lowest to highest, 30 to 60 mg, 15 to 30 mg, and PBO at study end. Dosing must be refined. The results provide support for additional examination of the agonist model in psychostimulant-dependence treatment.