Energy expenditure and the circulating concentration of various intermediary metabolites, insulin and glucagon, were measured in five lean subjects at rest and during a 20-min period of a standardized exercise (50-75 watts). Measurements were made before starvation, at the end of a 4-d period of total starvation and 24-32 h after refeeding. The respiratory quotient decreased in all subjects during starvation from 0.85 +/- 0.03 (s.e.m.) to 0.70 +/- 0.01 (P less than 0.01), and rose again on refeeding to 0.85 +/- 0.04. Resting metabolic rate (RMR) was not significantly affected by starvation. 'Work efficiency' (mechanical work done X 100 divided by metabolic rate during work - RMR) decreased in all subjects from a mean value of 23.9 per cent before starvation to 22.2 per cent during starvation and rose again on refeeding to 23.9 per cent, but with small numbers these differences did not reach statistical significance. All subjects felt that the work load (assessed on the Borg scale for perceived exertion) was greater during starvation than either before or after starvation (P less than 0.01). During exercise the circulating concentrations of glucose and glucagon remained virtually unchanged whereas insulin tended to decrease. In contrast, concentrations of lactate, pyruvate and alanine increased. The changes in the concentration of lactate, pyruvate and alanine were greater during starvation than before starvation, and are consistent with inhibition of the pyruvate dehydrogenase complex by ketone bodies, the circulating concentrations of which were elevated 20-fold during starvation. It is suggested that this inhibition may increase glucose recycling between muscle and liver and cause a small increase in energy expenditure.