One hundred and twenty-nine children, 2 to 8 years old, with idiopathic autistic spectrum disorder diagnosed by standard instruments (Childhood Austim Ratings Scale and Autism Diagnostic Observation Schedule) were treated with fluoxetine (0.15 to 0.5mg/kg) for 5 to 76 months (mean 32 to 36 months), with discontinuation trials. Response criteria are described. Family histories were obtained using the family history method in repeated interviews. Fluoxetine response, family history of major affective disorder, and unusual intellectual achievement, pretreatment language, and hyperlexia were used to define a coherent subgroup of autistic spectrum disorder. Statistical analyses were post hoc. Of the children, 22 (17%) had an excellent response, 67 (52%) good, and 40 (31%) fair/poor. Treatment age did not correlate with response. Fluoxetine response correlated robustly with familial major affective disorder and unusual intellectual achievement, and with hyperlexia in the child. Family history of bipolar disorder and of unusual intellectual achievement correlated strongly. Five children developed bipolar disorder during follow-up. Fluoxetine response, family history of major affective disorder (especially bipolar), unusual achievement, and hyperlexia in the children appear to define a homogeneous autistic subgroup. Bipolar disorder, unusual intellectual achievement, and autistic spectrum disorders cluster strongly in families and may share genetic determinants.