The value of 'whole body' and segmental impedance measurements, and of simple anthropometric methods for predicting body composition was assessed in 24 normal (14m, 10f) subjects (BMI, 18.3-28.6), using densitometry as the reference method. The contribution of segmental impedance was assessed in a separate group of 24 normal (12m, 12f) subjects (BMI, 19.8-28.8) at two frequencies (1 kHz and 50 kHz). Estimates of specific resistivities of certain individual segments (upper arm, forearm, upper leg, and lower leg) were also made in this group, and compared to those obtained from a group of 7 obese female subjects (BMI, 32.6-56.1). The bias and 95 per cent limits of agreement between densitometrically determined body composition (fat and fat-free mass, and total body water) and the alternative methods were found to vary considerably, depending on the technique and/or equations employed. Estimates of whole body composition based on impedance or resistance measurements were found to be associated with only slightly smaller limits of agreement than those made by anthropometry. The upper limb was found to have the greatest influence on whole body impedance measurements. Indeed, the forearm, which accounts for 1.3 per cent of body weight contributes 25.0 per cent to 'whole body' impedance. The estimated specific resistivities of segments were found to be considerably greater in the obese individuals than in normal female subjects (for example, 75 per cent higher for the upper arm, P less than 0.001). The results suggest that: (a) there may be a systematic, population-related, error in predicting densitometric estimates of body composition with the use of standard equations, which incorporate variables such as weight, height, skinfold thicknesses, and impedance/resistance measurements; (b) in this population, impedance or resistance measurements confer only a small advantage over simple anthropometry for predicting body composition; (c) the impedance of the arm or leg may provide a simple alternative method for assessing the composition of the whole body; and (d) the estimated specific resistivity of individual body segments may be useful for assessing the composition of those segments.