In this study, the bioelectrical impedance analysis (BIA), skinfold thickness measurement (STM) and dual-energy X-ray absorptiometry (DXA), as a reference method, were compared with each other in the assessment of body composition in elderly (62-72-year-old) Finnish women (n=93). BIA had better agreement with DXA in the assessment of fat free mass (FFM, R2=0.70, Sres=2.1) and fat mass (FM, R2=0.93, Sres=2.3) than the STM (FFM, R2=0.62, Sres=2.4; FM, R2=0.89,Sres=2.8). There was quite a large variation in the estimates when different BIA prediction equations were used. The equation developed in this study, FFM (kg)=-128.06 + 1.85 x BMI-0.63 x weight + 1.07 x height - 0.03 x resistance +10.0 x waist-hip ratio, yielded a small and unbiased error (0.5 +/- 1.6 kg), with a small residual standard deviation (R2=0.83, Sres=1.6). However, error associated with the estimate of FM was positively related to the degree of FM. BIA(Heitmann) equation yielded unbiased estimates of both FFM and FM (FFM, R2=0.77, Sres=1.8; FM, R2=0.95, Sres=1.9). Both the STM and BIA (manufacturer's equation) resulted in error which was statistically significant and positively correlated with FFM and FM. These results indicate that BIA prediction equations, chosen with care, can improve the performance of equations based upon anthropometric measurements alone in the assessment of body composition in elderly women.