We postulated that the commonly observed constant linear relationship between VO2 and work rate during cycle ergometry to exhaustion is fortuitous and not due to an unchanging cost of external work. Therefore we measured VO2 continuously in 10 healthy men during such exercise while varying the rate of work incrementation and analyzed by linear regression techniques the relationship between VO2 and work rate (delta VO2/delta wr). After excluding the first and last portions of each test we found the mean +/- SD of the delta VO2/delta wr in ml.min-1.W-1 to be 11.2 +/- 0.15, 10.2 +/- 0.16, and 8.8 +/- 0.15 for the 15, 30, and 60 W.min-1 tests, respectively, expressed as ml.J-1 the values were 0.187 +/- 0.0025, 0.170 +/- 0.0027 and 0.147 +/- 0.0025. The slopes of the lower halves of the 15 and 30 W.min-1 tests were 9.9 +/- 0.2 ml.min-1.W-1 similar to the values for aerobic work reported by others. However the upper halves of the 15, 30, and 60 W.min-1 tests demonstrated significant differences: 12.4 +/- 0.36 vs 10.5 +/- 0.31 vs 8.7 +/- 0.23 ml.min-1.W-1 respectively. We postulate that these systematic differences are due to two opposing influences: 1) the fraction of energy from anaerobic sources is larger in the brief 60 W.min-1 tests and 2) the increased energy requirement per W of heavy work is evident especially in the long 15 W.min-1 tests.