Neural tube defects remain the most serious common birth defect and, despite considerable progress in understanding these malformations, the etiology of most cases remains unknown. It has been proposed that the cause may vary with the type and location of the malformation but, if these variables are to be studied, a rigorous classification of cases is required. This has become more important as birth prevalence has fallen, mainly due to prenatal diagnosis and elective termination of pregnancy, and future studies will increasingly require collaboration between centres. In this study we have combined data from Ottawa, Ontario, and Boston, Massachusetts, in an attempt to examine the effect of level of spina bifida on sib occurrence rates and the rates of associated malformations, and to compare the level of lesion when determined radiographically with that recorded on the clinical chart. Malformations appeared to be more frequent with thoracic spina bifida and were more often associated with additional vertebral anomalies. Significant differences were found between the upper level of lesion recorded in the clinical file and that visible radiographically. Sib recurrences were too few for statistical comparison, but the data suggest a higher rate among sibs and more distant relatives of propositi with upper level lesions. There was not evidence to support a greater than expected concordance for level of lesion between sibs.