Evaluated were programs serving 43 children in 11 elementary classes for autistic children in California. Children were videotaped in three instructional settings: interaction with peers, interaction with adult, and a language sample session. Behavioral observations were completed on each child; records were examined for pertinent diagnostic data and individual educational plans; and personnel were interviewed regarding program philosophy. Observational data resulted in approximately 1/5 of the children being judged as "not autistic". Serious deficiencies were found in the availability and relevance of diagnostic data. Program philosophy tended to be heavily behavioral. Recommendations included establishing diagnostic criteria for determination of autism and including precise, measurable educational objectives in the individual educational plans. (DB)