Five men were studied during exercise to exhaustion on an electrically braked cycle ergometer at 70% of VO2max. The four experimental treatments were as follows: fasted for 36 h (A); fasted (36 h) and refed with glucose (B) or glycerol (C); postabsorptive (overnight fast, D). In B and C the subjects were given a drink containing glucose or glycerol (1g per kg body weight) 45 min before starting exercise. A placebo drink was given 45 min before exercise on treatments A and D. Despite an increased availability of circulating free fatty acids, beta-hydroxybutyrate and glycerol exercise time to exhaustion was significantly lower after fasting (treatment A 77.7 +/- 6.8 min) compared with treatment D (119.5 +/- 5.8 min). Refeeding with glucose or glycerol did not significantly improve performance (92.4 +/- 11.8 min and 80.8 +/- 3.6 min respectively) compared with treatment A and lowered circulating levels of FFA and beta-HB during exercise compared with A. Despite the probability of low liver glycogen levels after fasting, none of the subjects became hypoglycaemic (blood glucose less than 4 mmol.l-1) during exercise and their blood lactate concentrations were not high at exhaustion. Plasma levels of branched chain amino acids (BCAA) decreased progressively during exercise on treatments A, B and C and were considerably lower at exhaustion compared with treatment D. Falling plasma concentrations of BCAA during prolonged exercise may be implicated in the generation of central fatigue.