The aim of this study was to examine the effects of fatigue induced by a 65-km ultramarathon on the oxygen cost of running (Cr) and cycling (Ccycl). The day before and immediately after the race, a group of nine well-trained male subjects performed two submaximal 4-min exercise bouts: one cycling at a power corresponding to 1.5 W x kg(-1) body mass on an electromagnetically braked ergometer, and one running at 11 km x h(-1) on a flat asphalt roadway. Before oxygen cost determinations, the subjects performed 12 "ankle" jumps at a given frequency that was fixed by an electronic metronome (2.5 Hz). From the non-fatigued to the fatigued condition, there was a significant increase in minute ventilation for both running (P < 0.01) and cycling (P < 0.0001). Significant changes were also found in respiratory exchange ratio both for running (P = 0.01) and cycling (P < 0.0001). However, running and cycling differed in that Cycyc increased significantly by [mean (SD)] 24.2 (11.5)% (P < 0.001), suggesting an alteration of muscle efficiency, while Cr did not change with fatigue [186.8 (14.1) mlO2 x kg(-1) x km(-1) vs 186.8 (18.7) mlO2 x kg(-1) x km(-1)]. In addition, contact times during hopping increased significantly from 0.173 (0.019) ms to 0.194 (0.027) ms (P < 0.01). Analysis of the factors that determine Cr indicate that the subjects modified their movement pattern in order to decrease the mechanical cost of running in such long-term fatigue conditions.