There is a paucity of data on differences between methods for the assessment of body composition in elderly subjects. Studies on younger adults suggest that such differences are of some practical significance at the individual level. In the present study the following methods of estimating percentage body fatness (BF%) were compared in healthy elderly men and women (mean age 70 (SD 6) years: densitometry; skinfold thickness; total body water; bioelectrical impedance (BIA) using an age-specific predictive equation and the manufacturers' equation; body mass index (BMI). Though BF% estimates from the various methods tended to be highly correlated with those from densitometry and with each other, differences between methods at the individual level were marked. In particular, the age-specific equations based on BMI and BIA systematically overestimated BF% relative to the other methods. Biases between BF% estimates derived from densitometry, skinfolds, BIA (manufacturers' equation) and total body water were less marked, indicating little evidence of systematic differences between these methods in elderly subjects. Individual differences between methods were slightly greater than those reported in some studies of younger adults, but this may be of little practical significance, and may be considered inevitable in view of variability between and within subjects in the extent to which the underlying assumptions of these two-component methods are met in elderly subjects.