Thirty-seven children, aged between 2 and 7 years, with idiopathic autism underwent an open-label trial of fluoxetine treatment. All had assessment of diagnosis, developmental status, and family psychiatric history. Independent developmental testing before and after starting fluoxetine permitted quantification of language acquisition in a subgroup. Twenty-two of the 37 children had a beneficial treatment response sustained during continuing treatment for 13 to 33 months (mean 21 months). Eleven had an excellent response and were able to attend mainstream classrooms. Eleven had a good response though they remained identifiably autistic. Fifteen children had no benefit. Responders showed behavioral, language, cognitive, affective, and social improvements. Responders with adequate testing showed marked increases in language acquisition at every stage of development as compared with (1) pretreatment status, (2) responses to other treatments, (3) ability in non-language (matching) tasks, and (4) historical controls from the literature. The response to fluoxetine strongly correlated with a family history of major affective disorder. These preliminary findings implicate serotonergic mechanisms in autistic symptomatology and warrant further study with controlled trials.