Seven physically untrained subjects underwent 72 h total sleep deprivation, following a baseline day. Daily, at 0400 and 1600 h, subjects pedalled on a bicycle ergometer under individually set work loads of 40, 60, and 80% Vo2max: This was not a study oriented towards endurance but towards capacity, requinng steady-state measurement. From assessments 9fheart rate, VO2 and VCO2 were calculated: VO2max' gross mechanical efficiency, VO2 at a heart rate of 150, and respiratory quotient. To assess possible training effects, a control group underwent identical procedures except that they slept at night and had the morning measure delayed until 0830 h. A series of statistical models were applied to the data, which centered on quantifying the inherent underlying variability, to estimate the level any main effect had to reach to become significant. The analysis showed that the noise level was small enough for any real effect of importance to have been detected, with a reasonably large probability. No statistically significant effects were found for any of the parameters with respect to conditions, days, and time. The main significant outcome was with mechanical efficiency, which displayed greater variability during sleep deprivation. Both groups displayed similar trends in training effects. It was concluded that the physiological ability to do work of the type and duration used here was not adversely affected by 72 h of sleep loss.