The study was designed to evaluate the range of body composition in elite heavyweight oarswomen as well as the level of agreement between various methods used to measure this variable. Percent body fat was determined at the start of the competitive season by densitometry, taken to represent the reference standard, and measurement of total body potassium, skinfold thicknesses, bioelectrical impedance analysis and body mass index. The athletes were studied within a two week period with all measurements in any individual taken during one morning. We demonstrated a surprisingly large range of percent fat between these oarswomen, 13.6 to 29.3% by densitometry, which was a feature common to all methods. Percent body fat by total body potassium was lower (p < 0.05) while from body mass index higher (p < 0.01) than the reference value from densitometry. Similar methodologies generated significantly different estimates of % fat (SFT1 versus SFT2, p < 0.01 and BIAv versus BIAB, p < 0.01) highlighting the potential problems that may arise with the use of different regression equations to convert primary measurements into % fat. The limits of agreement between various methods were wide and reflect the large variability about the estimated mean bias. Practically this negates the correction of "non reference" values by adding or subtracting the mean difference or bias between the techniques in individuals. These methodological problems need to be considered when setting specific body composition targets for an athlete.