Epidemiologic data that were obtained through the SEER registry (Surveillance Epidemiology End Results Program) are presented. Survival statistics for five major brain tumor types in childhood are presented for the years 1968 through 1979. Further survival statistics are compared in patients treated in community hospitals versus university hospitals. The lack of uniform reporting, absence of pathologic conformation, and limited number of patients seen throughout the country are emphasized. There is a need for the establishment of a data base composed of the member institutions of the childhood cancer consortiums. This data base would address questions of patterns of failure, benefits of surgical and clinical staging, and the value of a new putative pathologic grading system. Additionally, the opportunity to collect these patients should permit identification of long-term treatment effects. Identification of early morbidity may lead to modification of treatment programs. A data base containing treatment and diagnostic parameters would allow significant cross-correlations and would lead to the design of future studies that are based upon accurate data.