Despite a high level of concern among health professionals and the general public, data are inadequate to define and monitor the prevalence of severe pediatric undernutrition (SPUN). The Centers for Disease Control is supporting four state health departments to test the feasibility of surveillance mechanisms for SPUN. These efforts will seek to develop workable case definitions, define prevalence estimates, identify high-risk populations and describe specific demographic, social and medical risk factors. Initial experience with SPUN surveillance indicates that a variety of approaches may prove feasible but that these require attention to issues such as measurement error, data collection burden, confidentiality and data management. SPUN surveillance may prove too costly to be practical for general application, but it can serve as a means to identify needy children and estimate the prevalence of undernutrition in specific high-risk populations.