A method is proposed for estimating the respiratory quotients (RQ) required in the calculation of free-living energy expenditure measured by the doubly-labelled water technique in man. Worked examples show that, in most normal subjects, measured or predicted food quotients (FQ) can be used in place of RQs since energy balance is usually maintained over the 10-20-d periods of double-isotope measurements. Examples of observed FQs in the UK are: omnivorous adults, 0.845 +/- 0.013 (s.d.); vegetarians, vegans and Asian immigrants, 0.860-0.880; breast-fed infants, 0.835 rising to 0.870 as weaning progresses; bottle-fed infants, 0.840-0.880. Alcohol intakes in excess of 2-3 per cent of total energy lower the FQ value. In most communities in the developing world FQs are substantially higher (0.900-0.955) due to the low contribution of fat to overall energy intakes; but FQs decrease progressively as the diet becomes more westernized. In the UK the between- and within-subject coefficients of variation based on 4-d weighed intakes are only 1.5 and 0.7 per cent respectively. A single 4-d measurement of dietary composition can therefore be used to predict a subject's FQ. In subjects in energy imbalance (eg, during growth, illness or when dieting) errors in calculated energy expenditure will rarely exceed 3-5 per cent even if the imbalance is ignored; in practice anabolism or catabolism can be accounted for and the FQ adjusted when converting FQ to RQ. The error incurred due to the substitution of adjusted FQ for RQ in the doubly-labelled water method will usually be negligible and should never exceed +/- 2 per cent.