This study compared body composition by dual-energy X-ray absorptiometry (DEXA; Lunar DPX-L) with that via a four-compartment (4C; water, bone mineral mass, fat, and residual) model. Relative body fat was determined for 152 healthy adults [30.0 +/- 11.1 (SD) yr; 75.10 +/- 14.88 kg; 176.3 +/- 8.7 cm] aged from 18 to 59 yr. The 4C approach [20.7% body fat (%BF)] resulted in a significantly (P < 0.001) higher mean %BF compared with DEXA (18.9% BF), with intraindividual variations ranging from -2.6 to 7.3% BF. Linear regression and a Bland and Altman plot demonstrated the tendency for DEXA to progressively underestimate the %BF of leaner individuals compared with the criterion 4C model (4C %BF = 0.862 x DEXA %BF + 4.417; r(2) = 0.952, standard error of estimate = 1.6% BF). This bias was not attributable to variations in fat-free mass hydration but may have been due to beam-hardening errors that resulted from differences in anterior-posterior tissue thickness.