This study examined the exercise responses of well-trained endurance athletes to various doses of caffeine to evaluate the impact of the drug on exercise metabolism and endurance capacity. Subjects (n = 8) withdrew from all dietary sources of caffeine for 48 h before each of four tests. One hour before exercise they ingested capsules of placebo or caffeine (3, 6, or 9 mg/kg), rested quietly, and then ran at 85% of maximal O2 consumption to voluntary exhaustion. Blood samples for methylxanthine, catecholamine, glucose, lactate, free fatty acid, and glycerol analyses were taken every 15 min. Plasma caffeine concentration increased with each dose (P < 0.05). Its major metabolite, paraxanthine, did not increase between the 6 and 9 mg/kg doses, suggesting that hepatic caffeine metabolism was saturated. Endurance was enhanced with both 3 and 6 mg/kg of caffeine (increases of 22 +/- 9 and 22 +/- 7%, respectively; both P < 0.05) over the placebo time of 49.4 +/- 4.2 min, whereas there was no significant effect with 9 mg/kg of caffeine. In contrast, plasma epinephrine was not increased with 3 mg/kg of caffeine but was greater with the higher doses (P < 0.05). Similarly only the highest dose of caffeine resulted in increases in glycerol and free fatty acids (P < 0.05). Thus the highest dose had the greatest effect on epinephrine and blood-borne metabolites yet had the least effect on performance. The lowest dose had little or no effect on epinephrine and metabolites but did have an ergogenic effect. These results are not compatible with the traditional theory that caffeine mediates its ergogenic effect via enhanced catecholamines.