Body fat content of seven lean women (body mass index (BMI) 20.6 (SD 1.8) kg/m2) and seven overweight women (BMI 31.1 (SD 3.3) kg/m2) was estimated by six different methods: underwater weighing (UWW), body-water dilution (BWD), whole-body counting (40K), skinfold thickness (SFT), bioelectrical impedance (BEI) and magnetic resonance imaging (MRI). Using UWW as the reference method, the differences between percentage fat by each other method and the percentage fat by UWW were calculated for each subject. The mean difference was lowest for SFT and highest for BWD. MRI showed the lowest variability in individual results, and 40K the highest. 40K and BWD methods used in combination gave better agreement with UWW results than either 40K or BWD methods alone. There was a weak negative correlation between the difference from the UWW results and percentage fat in the SFT measurements, but not in the BWD, 40K, BEI or MRI measurements, suggesting that for these methods the assumptions involved produced no greater inaccuracy in the overweight women than in the lean women. In all subjects the BEI offered little improvement over the traditional SFT measurements. The agreement between MRI and UWW estimates in both lean and overweight women suggests that MRI may be a satisfactory substitute for the more established methods of body fat estimation in adult women.