The ability of bioimpedance (BIA) to predict body composition in comparison with anthropometric measurements (weight and height) was assessed on three groups of adult young women (n = 99) and one group of adult young men (n = 49). Body fat (BF) and fat-free mass (FFM) by densitometry were used as the reference data. Resistance and reactance separately or together were poor predictors of BF and FFM, explaining from 0 to a maximum of 21 per cent of the FFM variation in the different groups. BF followed the same pattern, though the percentage of variance explained by both variables was even lower. Height squared divided by resistance (H2/R) explained from 22 to 68 per cent of the FFM variation and from 0 to 40 per cent of BF variation. Height alone was comparable to H2/R explaining from 11 to 53 per cent of the FFM variance in the four groups studied. Body weight was found to be the best single predictor of body composition; it explained from 56 to 78 per cent of FFM and 37 to 82 per cent of BF variability. Using stepwise regression analysis with all women combined, weight accounted for 70 per cent of the total FFM variation, with height and H2/R contributing only another 5 per cent. The same was found in men (68 vs 73 per cent respectively). The reported equation of Segal et al. was applied to our group, yielding almost the same high FFM prediction (r2 greater than 0.7 and SEE less than 2.5 kg).(ABSTRACT TRUNCATED AT 250 WORDS)