We examined the effect of caffeine co-ingested with either carbohydrate or fat on metabolism and performance in eight endurance-trained subjects who performed a random order of four experimental trials consisting of 120 min of steady-state ergometer cycling at 70 % of maximal O(2) uptake (SS) followed by a time trial in which subjects completed a set amount of work (7 kJ kg-1) as quickly as possible. One hour before SS subjects ingested either 2.6 g kg-1 carbohydrate (CHO); 2.6 g kg-1 CHO + 6 mg kg-1 caffeine (CHO + CAF); 1.2 g kg-1 fat with 2000 U I.V. heparin (FAT); or 1.2 g kg-1 fat with 2000 U I.V. heparin + 6 mg kg-1 caffeine (FAT + CAF). The rate of carbohydrate oxidation was higher (micromol kg-1 min-1: CHO, 243 +/- 39 and CHO + CAF, 239 +/- 30 vs. FAT, 196 +/- 48 and FAT + CAF, 191 +/- 55; P < 0.05, values are means +/- S.D.) and the rate of fat oxidation lower (micromol kg-1 min-1: CHO, 19 +/- 8 and CHO + CAF, 22 +/- 7 vs. FAT, 35 +/- 19 and FAT + CAF, 37 +/- 17; P < 0.05) with carbohydrate than fat ingestion. Yet despite lower carbohydrate use with fat feeding, the time taken to complete the time trial was less after carbohydrate than after fat ingestion (min: CHO, 30.37 +/- 7.42 and CHO + CAF, 29.12 +/- 5.62 vs. FAT, 33.02 +/- 8.50 and FAT + CAF, 32.78 +/- 7.70; P < 0.05). We conclude that (1) caffeine co-ingested with either carbohydrate or fat meals has no additive effect on substrate utilization or exercise performance and (2) carbohydrate ingestion before exercise improves subsequent time trial performance compared with fat ingestion. Experimental Physiology (2001) 86.1, 137-144.