This article discusses the use of epidemiologic and statistical approaches in developing definitions of obesity for children and adolescents. Definitions of adult obesity have been derived from mortality data or from statistical data on reference populations. Both approaches pose problems when applied to children and adolescents. The choice of appropriate measurements can be difficult. Data on childhood weight and adult outcomes are sparse and hard to interpret. Statistical definitions based on reference populations are more straightforward, but require arbitrary assumptions that limit their usefulness for prevalence estimates or comparisons across age, sex, or race/ethnic groups. Examples illustrate some issues in defining childhood obesity, including a case study of two groups of researchers who used the same data sets, but arrived at different conclusions. At present there is no generally accepted objective definition of obesity for children or adolescents.