The Autism Diagnostic Observation Schedule (ADOS), a standardized protocol for observation of social and communicative behavior associated with autism, is described. The instrument consists of a series of structured and semistructured presses for interaction, accompanied by coding, of specific target behaviors associated with particular tasks and by general ratings of the quality of behaviors. Interrater reliability for five raters exceeded weighted kappas of .55 for each item and each pair of raters for matched samples of 15 to 40 autistic and nonautistic, mildly mentally handicapped children (M IQ=59) between the ages of 6 and 18 years. Test-retest reliability was adequate. Further analyses compared these groups to two additional samples of autistic and nonautistic subjects with normal intelligence (M IQ=95), matched for sex and chronological age. Analyses yielded clear diagnostic differences in general ratings of social behavior, specific aspects of communication, and restricted or stereotypic behaviors and interests. Clinical guidelines for the diagnosis of autism in the draft version of ICD-10 were operationalized in terms of abnormalities on specific ADOS items. An algorithm based on these items was shown to have high reliability and discriminant validity. (author abstract)