The hypothesis that specific computerized tomography brain-scan findings are associated with infantile autism was tested in 45 cases and 19 controls. The autistic group was subdivided into serious and less-serious language-impaired subgroups. The analysis of Euclidean Distances, a type of cluster analysis, showed that neuroradiological parameters of cases and controls, including ventricular sizes, were on the whole significantly different, but no statistically significant difference appeared between the two autistic subgroups. But the analysis of variance of each neuroradiological parameter did not show any significant difference between autistics and controls. It was concluded that autism is nonspecifically associated with brain-scan abnormalities, and that other nonorganic, as well as organic, factors should be taken into account.