Body composition was measured in nine healthy, normal-weight, weight-stable subjects in three different research centres. In each centre the usual procedures for the measurements were followed. It revealed that the measurement procedures in the three centres were comparable. Body composition was measured in each centre between 09.00 and 13.00 hours after a light breakfast by densitometry (underwater weighing) and bio-electrical impedance. A single, total-body-water determination by D2O dilution was used as a reference value. Body fat determined by densitometry was significantly lower in one centre, which, however, could be completely explained by a lower body weight, probably due to water loss (the subjects refrained for a longer time from food and drinks before the measurements in that centre) and, thus, by violation of the assumptions of Siri's (1961) formula. Also, body impedance was slightly higher in that centre, indicating a lower amount of body water. Mean body fat from densitometry was also slightly lower in that centre compared with body fat determined by D2O dilution. Individual differences between body fat from densitometry and from total body water were relatively large, up to 7% body fat. The relationship between fat-free mass from densitometry and bio-electrical impedance was not different between the centres. It is concluded that differences in the relationship between body composition and bio-electrical impedance, as reported in the literature, may be due to differences in standardization procedures and/or differences in reference population.