The Tanita body-fat analyser is a novel device to estimate body fat, based on the principles of bioelectrical impedance. It differs from other impedance systems which use surface electrodes in that the subjects stand bare-footed on a metal sole-plate which incorporates the electrodes, hence impedance is measured through the legs and lower trunk. In 104 men and 101 women (16–78 years and BMI 16–41 kg/m2) the mean bias in body-fat mass measured using the Tanita body-fat analyser was 0·8 (2SD 7·9) KG RELATIVE TO A FOUR-COMPARTMENT MODEL. THIS IS COMPARABLE TO THE OTHER PREDICTION TECHNIQUES TESTED (CONVENTIONAL TETRAPOLAR IMPEDANCE -1·3 (2sd 6·9) kg, skinfold thicknesses 0·3 (2sd 7·4) kg, and BMI-based formulas -0·2 (2sd 9·0) kg and -0·6 (2sd 8·5) kg), but the agreement was poorer than for ‘reference’ methods to measure body fat (density 0·2 (2sd 3·7) kg, total body water -0·9 (2sd 3·4) kg and dual-energy X-ray absorptiometry 0·1 (2sd 5·0) kg). The present paper also describes the derivation of a new prediction equation for the calculation of body composition from the Tanita body-fat analyser. The equation incorporates sex, age, and a log-transformation of height, weight and the measured impedance to predict body fat measured by a four-compartment model. This approach is recommended in the derivation of other prediction equations in body composition analysis. Using this novel prediction equation the residual standard deviations were 4·8 % for men and 3·3 % for women. A similar analysis using data collected with a conventional tetrapolar system yielded residual standard deviations of 4·3 % for men and 3·1 % for women. This demonstrates that the practical simplicity of the novel Tanita method is not associated with a clinically significant decrement in performance relative to a traditional impedance device.