The concept of the accumulated O-2 deficit (AOD) assumes that the O-2 deficit increases monotonically with increasing work rate (WR), to plateau at the maximum AOD, and is based on linear extrapolation of the relationship between measured steady-state oxygen uptake ((V) over dot O-2) and WR for moderate exercise. However, for high WRs, the measured (V) over dot O-2 increases above that expected from such linear extrapolation, reflecting the superimposition of a 'slow component' on the fundamental (V) over dot O-2 mono-exponential kinetics. We were therefore interested in determining the effect of the (V) over dot O-2 slow component on the computed AOD. Ten subjects [31 (12) years] performed square-wave cycle ergometry of moderate (40%, 60%, 80% and 90% theta(L)) heavy (40%Delta), very heavy (80%Delta) and severe (110% (V) over dot O-2peak) intensities for 10-15 min, where (theta) over capL is the estimated lactate threshold and Delta is the WR difference between (theta) over cap (L) and (V) over dot O-2peak. (V) over dot O-2 was determined breath-by-breath. Projected 'steady-state' (V) over dot O-2 values were determined from sub-(theta) over capL tests. The measured (V) over dot O-2 exceeded the projected value after similar to3 min for both heavy and very heavy intensity exercise. This led to the AOD actually becoming negative. Thus, for heavy exercise, while the AOD was positive [0.63 (0.41) l] at 5 min, it was negative by 10 min [-0.61 (1.05) l], and more so by 15 min [-1.70 (1.64) l]. For the very heavy WRs, the AOD was [0.42 (0.67) l] by 5 min and reached -2.68 (2.09) l at exhaustion. For severe exercise, however, the AOD at exhaustion was positive in each case: +1.69 (0.39) l. We therefore conclude that the assumptions underlying the computation of the AOD are invalid for heavy and very heavy cycle ergometry (at least). Physiological inferences, such as the 'anaerobic work capacity', are therefore prone to misinterpretation.