The lack of chemical maturity, subsequent to normal growth and maturation, complicates the assessment of body composition in children. Even though known to overpredict percent body fat (%BF), the use of adult prediction equations is widespread in the pediatric literature. Many sex- and age-specific modified equations have been proposed. This study reports a cross-validation analysis of selected laboratory-based criterion methods in an attempt to identify the most appropriate reference which could be used in future validation studies of the more practical field/clinical testing methods. Subjects were 48 (24 boys and 24 girls) peripubertal children evenly distributed according to stage of maturation. Criterion measurements included body density (Db) by hydrostatic weighing, total body water (TBW) by deuterium oxide dilution, and total body mineral content (TMC) by dual energy x-ray absorptiometry. Five different prediction models were evaluated. Test-retest reliabilty was high (ICC = .970 to .999). Of the prediction models tested, the four-component model was considered the most accurate laboratory-based criterion model since it involves measurement of the primary constituents of fat-free mass. Based on high r (≥.942) values, low standard errors of estimate (SEE = 1.8 %BF males, 1.1 %BF females); and low total prediction errors (TE = 1.9 %BF males, 2.0 %BF females), the Lohman age-adjusted prediction equation showed the best agreement with the four component model. If multiple testing facilities are unavailable, the Lohman two-component (Db) model would be the criterion method of choice. © 1993 Wiley-Liss, Inc.