To evaluate the prevalence of von Willebrand's disease (vWd) we carried out an epidemiological investigation among school children of the Veneto region in northern Italy. A total of 1,218 of 1,281 possible children participated in the study. They were 11 to 14 years of age, and all attended secondary schools in two distinct small areas, 70 km apart, between which there is no social contact. A blood sample was taken from each subject for determination of the blood group and von Willebrand factor (vWf) level (measured as ristocetin cofactor and expressed in IU/dL after calibration of the internal pool against an international standard), and the parents were given a questionnaire concerning hemorrhagic symptoms in the members of the family in the last three generations. Separate normal ranges were calculated for blood group O and non-O subjects (1,166 children and 289 adults) with a nonparametric method because the distribution curves of the reference values did not fit the gaussian distribution. Diagnoses of vWd were considered only for children who had low vWf levels and were members of a family with a convincing bleeding history (case of "probable vWd"). A final diagnosis was assigned if, in addition to these criteria, at least one other family member on the side with hemorrhagic history had a low vWf level. Of the 1,218 children examined, ten were classified as having vWd (0.82%). Taking into account the 90% confidence interval for the lower limit of the normal range, this figure could range from 7 (0.57%) to 14 (1.15%). All these subjects were mildly to moderately affected and presented features of heterozygous classic vWd (type I). Affected subjects were distributed evenly in the two areas examined. Our results suggest that the prevalence of vWd might be much higher than previously reported and that a different screening approach might be of use for patients with mild bleeding diathesis.