Relationships between incidence of Wilms' tumour and information recorded at birth were investigated in a prospective study of the 1,489,297 children born in Norway between 1967 and 1992. A total of 119 individuals were diagnosed with Wilms' tumour in the age interval 0-14 years. A high length at birth was significantly associated with a high risk (incidence rate ratio 1.8 for length > or = 53 cm vs or = 9, 95% CI 1.2-3.9). For all variables for which an association was indicated, the association seemed to be restricted mainly to children aged less than 2 years. This suggests that Wilms' tumour diagnosed early in life may differ aetiologically from that of cases diagnosed later.