Six children with the diagnosis of Attention Deficit Disorder were treated as day hospital patients, using different stimulant medication. They were studied in a double-blind crossover experiment in which they received caffeine in low dose or in a high dose. Methylphenidate was added to both dosages, as well as administered alone. Results indicated that caffeine in low dosage when added to methylphenidate was superior to all other treatment conditions. Caffeine in low dosage could not be differentiated from 10 mg of methylphenidate. High dosage caffeine was not different from placebo or no-drug conditions. This study offers evidence to support a curvilinear pattern of dose-response for caffeine, in attenuating the behavioural manifestations of this syndrome.