The effects were compared of exercise in the fasted state and exercise with a high rate of carbohydrate intake on intramyocellular triglyceride (IMTG) and glycogen content of human muscle. Using a randomized crossover study design, nine young healthy volunteers participated in two experimental sessions with an interval of 3 weeks. In each session subjects performed 2 h of constant-load bicycle exercise (~75% V_O2,max), followed by 4 h of controlled recovery. On one occasion they exercised after an overnight fast (F), and on the other (CHO) they received carbohydrates before (~150 g) and during (1 g (kg bw)^-1 h^-1) exercise. In both conditions, subjects ingested 5 g carbohydrates per kg body weight during recovery. Fibre type-specific relative IMTG content was determined by Oil red O staining in needle biopsies from m. vastus lateralis before, immediately after and 4 h after exercise. During F but not during CHO, the exercise bout decreased IMTG content in type I fibres from 18 ± 2% to 6 ± 2% (P = 0.007) area lipid staining. Conversely, during recovery, IMTG in type I fibres decreased from 15 ± 2% to 10 ± 2% in CHO, but did not change in F. Neither exercise nor recovery changed IMTG in type IIa fibres in any experimental condition. Exercise-induced net glycogen breakdown was similar in F and CHO. However, compared with CHO (11.0 ± 7.8 mmol kg^-1 h^-1), mean rate of postexercise muscle glycogen resynthesis was 3-fold greater in F (32.9 ± 2.7 mmol kg^-1 h^-1, P = 0.01). Furthermore, oral glucose loading during recovery increased plasma insulin markedly more in F (+46.80 µU ml^-1) than in CHO (+14.63 µU ml^-1, P = 0.02). We conclude that IMTG breakdown during prolonged submaximal exercise in the fasted state takes place predominantly in type I fibres and that this breakdown is prevented in the CHO-fed state. Furthermore, facilitated glucose-induced insulin secretion may contribute to enhanced muscle glycogen resynthesis following exercise in the fasted state.