We hypothesised that experienced runners would select a stride frequency closer to the optimum (minimal energy costs) than would novice runners. In addition, we expected that optimal stride frequency could simply be determined by monitoring heart rate without measuring oxygen consumption (VO2). Ten healthy males (mean ±s: 24±2 year) with no running training experience and 10 trained runners of similar age ran at constant treadmill speed corresponding to 80% of individual ventilatory threshold. For two days, they ran at seven different stride frequencies (self-selected stride frequency ±18%) imposed by a metronome. Optimal stride frequency was based on the minimum of a second-order polynomial equation fitted through steady state VO2 at each stride frequency. Running cost (mean±s) at optimal stride frequency was higher (P < 0.05) in novice (236±31 ml O2·kg(-1.) km(-1)) than trained (189±13 ml O2·kg(-1.) km(-1)) runners. Self-selected stride frequency (mean ±s; strides(.)min(-1)) for novice (77.8±2.8) and trained runners (84.4±5.3) were lower (P < 0.05) than optimal stride frequency (respectively, 84.9±5.0 and 87.1±4.8). The difference between self-selected and optimal stride frequency was smaller (P < 0.05) for trained runners. In both the groups optimal stride frequency established with heart rate was not different (P > 0.3) from optimal stride frequency based on VO2. In each group and despite limited variation between participants, optimal stride frequencies derived from VO2 and heart rate were related (r > 0.7; P < 0.05). In conclusion, trained runners chose a stride frequency closer to the optimum for energy expenditure than novices. Heart rate could be used to establish optimal stride frequency.