1. Seven healthy young men of sedentary habits were submitted to a 10-week period of endurance physical training on a cycle ergometer. The training program caused a 15% increase in maximal oxygen consumption (VO2max) and a 16% reduction in resting heart rate (HR). Before and after training, these volunteers performed dynamic exercise (DE) on a cycle ergometer at loads of 25, 50, 75, 100 and 150 w for 4 min at each level. The same exercise protocol was applied to 13 sedentary individuals and to 7 athletes (medium distance runners) who showed a VO2max of 39.4 and 53.8 ml/kg, respectively. HR was continuously monitored throughout the period of effort at each workload. 2. During the first 10 s of DE, a period when tachycardia is mediated almost exclusively by vagal withdrawal, the athletes presented a more rapid increase in HR than sedentary subjects. The same tendency was observed in the sedentary individuals after the training period, although of a lesser magnitude. 3. During the DE phase in which sympathetic mediation plays an important role (between 30 s and 4 min), the athletes presented a lower HR increase than the sedentary individuals, and the same response pattern was observed in the group submitted to physical training. Total HR increase (from 0 to 4 min) induced by DE was lower in athletes than in sedentary subjects and was not changed by training of the sedentary subjects. 4. These results suggest that aerobic training decreases the slow sympathetic and increases the fast parasympathetic contribution to HR during dynamic exercise at the same absolute workloads. 5. These functional changes in the autonomic control of HR may or may not be associated with modifications of absolute HR values which increase from rest to the end of exercise. In contrast to what happens in athletes, the autonomic adaptations observed after short-term aerobic training may occur during DE without changes in the total HR response.