Animal models of human brain damage syndromes range from ones that are purely theoretical t o ones that attempt t o duplicate a human syndrome a t every level. The similarity between animal and human cases in experimental manipulation, CNS alteration, and tests of parallelism is sometimes restricted by the extent t o which experimental species match humans in their basic biology. However, the information available from the human condition is often the most important limiting factor in the degree t o which similarities can be achieved. The recent discovery of a high rate of autism in patients exposed t o thalidomide around the time of neural tube closure suggests an appropriate experimental manipulation for an animal model of the disease. Rats exposed t o valproic acid a t the same stage of development mimic the thalidomide cases' apparent injuries t o the brain stem. Whether the behavior of these braindamaged animals is similar t o that observed in autism is difficult t o determine, because the diagnostic behavioral features of the disease are not easily translated into animal tasks. However, anatomical observations in human cases provide tests of parallelism that can be applied t o animals, and these appear t o be positive in the animal model. The development of behavioral tests that discriminate "autistic behavior" from that of other brain damage syndromes may require further research in both humans and animals.