In several species, physical conditioning (training) provokes a large shift in substrate utilisation during submaximal exercise. Few studies in horses have quantitatively examined these effects. Therefore, the effects of exercise training on plasma glucose kinetics during submaximal exercise were examined in 7 horses (5 Thoroughbred, 2 Standardbred; age 3-9 years) that had been paddock-rested for at least 6 months. Two days after determination of maximum aerobic capacity (VO2max), horses ran on a treadmill (4 degree incline) at 55% of VO2max (UT) for 60 min or until fatigue and then completed 6 weeks of moderate-intensity training on a treadmill (5 days/week). Following training and a second VO2max test, the horses completed exercise trials at the same absolute (ABS) and relative (REL) workload in random order, with at least 3 days between tests. After training, VO2max had increased (P<0.05) by 14.9% (mean +/- s.e. pretraining 118.4 +/- 7.4 ml/kg bwt/min; post-training 136.1 +/- 7.8 ml/kg bwt/min). Mean exercise duration was longer (P<0.05) in the ABS trial (57 +/- 1.9 min) than in the UT (46 +/- 3.9 min) and REL (49 +/- 4.6 min) trials. Plasma glucose concentration increased during exercise, and was lower (P<0.05) in ABS than in UT and REL at the end of exercise. Mean glucose rate of appearance (Ra) and disappearance (Rd) were 22 and 21% lower (P<0.05), respectively, in ABS than in UT, but mean glucose Ra and Rd did not differ between the UT and REL trials. Exercise-induced changes in glucagon, epinephrine and norepinephrine were blunted (P<0.05) in ABS, but not REL, when compared to UT. It is concluded that 6 weeks of moderate-intensity training results in a decrease in glucose flux during submaximal exercise at the same absolute, but not relative, workload. The training-induced decrease in glucose flux may, in part, be due to altered plasma concentrations of the major glucoregulatory hormones.