The purpose of this study was to examine the relationship between cadence and oxygen consumption with exercise duration. Ten triathletes who trained regularly were examined. The first test was always a maximal test to determine maximal oxygen uptake (VO2max). The other sessions were composed of six submaximal tests representing 80% of the maximal power reached with VO2max (Pmax). During these tests submaximal rides with a duration of 30 min were performed. Each test represented, in a randomised order, one of the following pedal rates: 50, 65, 80, 95, 110 rpm and a freely-chosen rate. VO2, respiratory parameters, and heart rate were monitored continuously. Two periods, between the 3rd and the 6th minute and between the 25th and the 28th minute, were analysed. Results showed that when VO2 and heart rate were plotted against cadence, each curve could be best described by a parabolic function, whatever the period. Furthermore, a significant effect of period was found on energetically optimal cadence (70 +/- 4.5 vs. 86 +/- 6.2 rpm, P < 0.05). Only during the second period was no significant difference found between freely-chosen cadence (83 +/- 6.9 rpm) and energetically optimal cadence (P > 0.05). In conclusion, our results suggest that during prolonged exercise triathletes choose a cadence that is close to the energetically optimal cadence. A change of muscle fibre recruitment pattern with exercise duration and cadence would explain the shift in energetically optimal rate towards a higher pedal rate observed at the end of exercise.