An expert panel reviewed a cluster of childhood leukemias in Fallon, Nevada, and suggested the population mixing hypothesis as an explanation. This hypothesis proposes that nonimmune children exposed to some unknown infection(s), through population mixing, are at increased risk of developing acute lymphoblastic leukemia and non-Hodgkin's lymphoma. The United Kingdom Childhood Cancer Study registered 3,838 children with cancer and 7,669 matched controls aged 0-14 years during 1991-1996 in England, Scotland, and Wales. Local area characteristics for each child's residential address at diagnosis were assigned from census data: volume and diversity of population mixing, material deprivation, and rural status. The best-fitting models were chosen for three diagnostic groups: acute lymphoblastic leukemia, non-Hodgkin's lymphoma, and all other tumors. Elevated risks of acute lymphoblastic leukemia were found in areas with a low diversity of origins of migrants and for non-Hodgkin's lymphoma in areas with a low diversity of origins of child migrants; for other tumors, no covariates were associated. This study, and a survey of 17 published reports on population mixing, suggests that a low diversity of migrant backgrounds may be associated with acute lymphoblastic leukemia. These findings do not support the population mixing hypothesis. Although they support the Greaves delayed infection hypothesis, other aspects of this hypothesis were not addressed.