Prenatal exposure to high levels of alcohol often induces birth defects that combine morphological stigmata with neurological or neuropsychological deficits. But it has proved problematic to diagnose these syndromes in adolescents and adults, in whom the morphological signs are absent or attenuated, the behavioral deficits nonspecific, and the exposure history often difficult to reconstruct. Localizing the associated brain abnormalities might circumvent most of these difficulties. To this end, three-dimensional (3D) locations were recorded for 67 homologous points on or near the corpus callosum in magnetic resonance (MR) brain images from 60 adolescents and adults who were normal, 60 diagnosed with fetal alcohol syndrome, and 60 diagnosed with fetal alcohol effects. We combined the standard statistical approach to this type of geometric data, Procrustes analysis, with a multivariate strategy focusing on differences in variability. In this data set, the shape of the corpus callosum and its vicinity proves systematically much more variable in the alcohol-affected brains than in those of the normal subjects. From this excess variability follows a promising classification rule, having both high sensitivity (100 out of 117) and high specificity (49 out of 60) in this sample. The discrimination uses four landmark points and two summary scores of callosal outline shape. The information from the corpus callosum and vicinity, as viewed in MR brain images of full-grown subjects, may serve as a permanent record of the prenatal effects of alcohol, even in patients who are first suspected of these syndromes relatively late in life or who lack the facial signs of prenatal alcohol damage. The statistical pattern underlying the callosal diagnosis also leads to speculations on mechanisms of the prenatal damage. Anat Rec (New Anat) 269:162–174, 2002. © 2002 Wiley-Liss, Inc.