This study compared the effects of carbohydrate ingestion throughout exercise with ingestion of an equal amount of carbohydrate late in exercise. Eight well-trained men cycled 2 h at 70 +/- 1% VO2 peak, followed immediately by a 15-min performance ride, while ingesting either a 7% carbohydrate-electrolyte solution (CHO-7), an artificially sweetened placebo (CON), or the placebo for the first 90 min then a 21% glucose solution (CHO-0/21). At the start of the performance ride, plasma glucose averaged 4.2 +/- 0.2, 5.2 +/- 0.1, and 5.7 +/- 0.2 mmol.l-1 in CON, CHO-7, and CHO-0/21, respectively (all different, P < 0.05). Plasma insulin levels were similar just prior to the performance ride in CHO-7 and CHO-0/21, with both higher than CON. A similar pattern was observed with respiratory exchange ratio (RER). Work performed during the performance ride was significantly greater in CHO-7 (268 +/- 8 kJ) compared with CON (242 +/- 9 kJ). Performance in CHO-0/21 (253 +/- 10 kJ), however, was not improved compared with CON, despite higher plasma glucose levels and plasma insulin levels similar to CHO-7. Seven of the eight subjects performed best in CHO-7. In conclusion, performance was improved, relative to the control trial, only when carbohydrate was ingested throughout exercise. Carbohydrate ingestion late in exercise did not improve performance despite increases in plasma glucose and insulin.