PURPOSE: This study was designed to compare the accuracy and bias in estimates of total body density (Db) by hydrostatic weighing (HW) and the BOD POD, and percent body fat (%fat) by the BOD POD with the four-compartment model (4C model) in 42 adult females. Furthermore, the role of the aqueous and mineral fractions in the estimation of body fat by the BOD POD was examined. METHODS: Total body water was determined by isotope dilution ((2)H(2)0) and bone mineral was determined by dual-energy x-ray absorptiometry. Db and %fat were determined by the BOD POD and HW. The 4C model of Baumgartner was used as the criterion measure of body fat. RESULTS: HW Db (1.0352 g x cm(-3)) was not statistically different (P = 0.35) from BOD POD Db (1.0349 g x cm(-3)). The regression between Db by HW and the BOD POD significantly deviated from the line of identity (Db by HW = 0.90 x Db by BOD POD + 0.099; R(2) = 0.94). BOD POD %fat (28.8%) was significantly lower (P < 0.01) than %fat by the 4C model (30.6%). The regression between %fat by the 4C model and the BOD POD significantly deviated from the line of identity (%fat by 4C model = 0.88 x %fat by BOD POD + 5.41%; R(2) = 0.92). BOD POD Db and %fat showed no bias across the range of fatness. Only the aqueous fraction of the fat-free mass (FFM) had a significant correlation with the difference in %fat between the 4C model and the BOD POD. CONCLUSION: These data indicate that the BOD POD underpredicted body fat as compared with the 4C model, and the aqueous fraction of the FFM had a significant effect on estimates of %fat by the BOD POD.