The purpose of this study was to identify the BIA (bioimpedance analysis) equation that yields the best estimate of body composition for 122 premenopausal African-American women (18-40 yr). Total body density (Db) was determined by hydrodensitometry at residual lung volume and converted to %BFHD using the Siri (31) formula, %BFHD was used to calculate reference fat-free mass (FFM). Resistance and reactance were measured using a Valhalla bioimpedance analyzer. The predictive accuracy of generalized, age-gender, race-specific, fatness-specific, and the Valhalla manufacturer's BIA equations was compared. There were significant correlations between FFMHD and FFMBIA for all BIA equations (r = 0.85 to 0.92). Except for the modified Segal fatness-specific equations, the prediction errors (SEE and E) exceeded 2.8 kg. For individuals, the %BF derived from FFMBIA predicted by the modified Segal equations was within +/- 3.5% BF for 69% of the subjects. This percentage was less (34-53%) for other equations. These results suggest that the predictive accuracy of BIA for estimating body composition of African-American women is improved when fatness-specific equations are used. We recommend using the modified Segal fatness-specific equations to assess FFM and %BF of premenopausal African-American women.