In recent years, researchers have focused attention on children who are exposed to domestic violence. Although presently there are no scientifically credible estimates of the national prevalence of children exposed to domestic violence, existing data suggest that large numbers of American children are affected. This article discusses the limitations of current databases and describes a promising model for the collection of reliable and valid prevalence data, the Spousal Assault Replication Program, which uses data collected through collaboration between police and university researchers.