The use of anthropometric measures (eg, skinfold thicknesses, BMI) to assess obesity is not without controversy and has not been explored with respect to the use among groups of children with growth retardation (ie, stunting). Therefore, the objective of this study was to determine whether growth retardation affects the accuracy of field methods for assessing body composition in children. A cross-sectional study was conducted in 30 stunted children and 30 nonstunted children who were matched for age- and weight-for-height z score and living in the shantytowns of São Paulo, Brazil. Body composition (fat mass, fat-free mass, and percentage of body fat [%BF]) was measured by H2(18)O dilution (reference technique) using group-specific values for the hydration of fat-free mass and dual-energy x-ray absorptiometry. BMI and body composition that were calculated from 3 pediatric skinfold prediction equations were evaluated for accuracy of %BF in comparison with the reference technique. Stunted children were shorter and weighed less than nonstunted children, but BMI did not differ significantly between groups. All 3 skinfold equations tested resulted in a calculated %BF that was significantly lower than that measured by H2(18)O dilution for both stunted and nonstunted groups, and %BF as calculated by any of the skinfold equations tested did not significantly predict %BF by H2(18)O dilution. In contrast, BMI significantly predicted %BF in both stunted and nonstunted children, and this relationship did not differ by growth status. BMI but not skinfolds significantly predicted %BF measured by H2(18)O dilution. The relationship between BMI and %BF did not differ between stunted and nonstunted children; this indicates that BMI can be used in field studies of obesity and stunting. However, the prediction of %BF by BMI is relatively poor in both groups of children, and continued investigation of more accurate field methods for measuring %BF is warranted.