A population-based study was carried out on 5790 tumours in children (aged 0-14 years) diagnosed in the period 1943-1984 in Denmark. Cases were identified from the files of the high-quality National Cancer Registry in which codes for tumours were based solely on topography until the end of 1977. To achieve a uniform data set following the outlines of the International Classification of Diseases for Oncology (ICD-O) coding system used by the Cancer Registry today, all cases of childhood cancer diagnosed prior to 1978 were re-evaluated, and an ICD-O code was applied. Tumours were aggregated into diagnostic groups suitable for analysis and presentation using an internationally agreed scheme, which was designed by the Manchester Children's Tumour Registry and modified recently by the International Agency for Research on Cancer. The average incidence rates for all histological types of childhood cancer combined were 137 per million boys and 111 per million girls, which are close to those reported from the USA but higher than most of the overall figures reported from Europe. The proportions of specific tumours were similar to those observed in other industrialized countries. The well known excess of cancer cases among boys compared to girls was due mainly to the occurrence of 90% more lymphomas, 30% more leukaemias and 15% more tumours of the central nervous system (CNS) among boys. Although significant increases were seen in the subgroups of CNS neoplasms and neuroblastomas (both sexes) and of lymphomas (boys only), no overall increase in childhood cancer was observed during the 42-year period of registration. While the increase in the incidence of CNS tumours was explained at least partly by better cancer surveillance, no interpretation can be offered for the increases seen for neuroblastomas and lymphomas. Our descriptive data suggest that environmental exposures do not play any significant role in the aetiology of the majority of childhood cancers.