Despite the recognition that autism is a common disorder with enormous public health significance, no medications have been shown to consistently improve the core social impairment of autism and available medications are most effective at reducing severe irritability and aggression. Combining two medications has proven effective in a number of other brain disorders, but has received little attention in autism. This study will investigate the hypothesis that memantine can improve certain aspects of autism including social withdrawal, inattention, and other behavioral manifestations of the disorder. We plan to test this hypothesis by treating children and adolescents who are already receiving atypical antipsychotics, and add memantine or placebo to assess the improvement of symptoms with memantine. The participants will be followed with standardized assessments and rating scales. Significance: Considering the clinical complexity of the syndrome and the well-characterized symptom domains, it may be necessary to treat individuals with one or more pharmacological and behavioral interventions. This study will be one of the first placebo-controlled trials of memantine in autism, as well as one of the first examining coactive medication treatment strategies for this difficult-to-treat disorder.