A group of 46 children in kindergarten through fourth grade whom their teachers considered to have severe behavior and/or learning disorders were randomly assigned to receive dextroamphetamine or placebo in a double-blind crossover design. D-amphetamine produced behavioral improvement that was clearly distinguishable from placebo effects in 25 children (54 percent of the study population). This represents seven percent of the total school population. The greatest improvement was seen in children with at least one hard neurological sign or two or more soft signs and a high hyperkinetic syndrome score. However, until a method of picking children with a bad prognosis is known, large-scale case finding in the schools with recommendation for long-term d-amphetamine treatment is not advised.