Based on neuroimaging and autopsy research, in autism no common site or type of abnormality appears in the cerebral hemispheres, thalamus, lenticular nuclei, and caudate nucleus. Nonetheless, further imaging and autopsy studies on this issue can be anticipated. Limbic system abnormalities have been reported at autopsy by one laboratory but not another, and no abnormality was found by the one quantitative MR study to measure a limbic structure. More autopsy and imaging research on the limbic system is needed. The cerebellum is the only anatomical structure for which there is both imaging and autopsy evidence of abnormality based on data gathered by many laboratories. Also, the only autopsy study to conduct statistical analyses of cerebellar cell loss found statistically significant Purkinje cell loss in both the vermis and hemispheres. Despite this, normal findings on routine radiologic examination are not of diagnostic significance at this time. On the one hand, the autopsy data show that most, if not all, autistic individuals have cerebellar cell loss, but on the other, research shows that MR images of the cerebellum in a substantial proportion of autistic individuals (perhaps 20% to 50%) will be indistinguishable from normal. Thus, it is likely that MR technology is not yet sufficiently sensitive to detect cerebellar abnormalities in all autistic persons who have them. Finally, the cerebellum seems an unlikely site of damage for a developmental disorder of higher cognition such as autism. However, new neurophysiologic and neuropsychologic studies of children with hemicerebellar resections and children with hemicerebellar resections and children with autism present an entirely new picture of the role of the cerebellum in normal human cognition in general and in the development of the social and communication deficits in autism in particular. These studies show that autistic subjects and patients with acquired cerebellar damage are unable to rapidly shift their mental focus of attention.