The FAO/WHO/UNU recommendations for energy requirements assume that the energy cost of sleep is equal to the basal metabolic rate (BMR). We have tested the validity of this assumption by analysing overnight and BMR measurements made by whole-body indirect calorimetry. Data from 80 healthy subjects measured on a total of 246 occasions have been used. In a subgroup of 40 normal lean subjects the mean ratio of overnight metabolic rate (Overnight MR): BMR was 0.95 (range 0.85 - 1.02, s.d. 0.04). The mean ratio of lowest sleeping metabolic rate (Lowest SMR): BMR was 0.88 (range 0.83 - 0.96, s.d. 0.04). Ratios of Overnight MR: BMR were not significantly affected by different levels of exercise on the preceding day. This ratio was significantly higher for subjects who were obese, late pregnant or attached to ECG electrodes. With the exception of the late pregnant subjects these groups had the same Lowest SMR:BMR ratios as the normal lean subjects, indicating that the higher Overnight MR was caused by disturbed sleep. The data suggest that the use of BMR to estimate overnight energy expenditure would introduce an average overestimate of approximately 5 per cent during the actual hours of sleep, but that when applied over 24 h the error becomes negligible.