1. At work rates which do not result in a sustained increase in blood lactate ([L-]), oxygen uptake (VO2) approaches the steady state with first-order kinetics. However, when [L-] is increased, at least two kinetic components are required to characterize the VO2 response dynamics. The purpose of the present investigation was to determine whether these more-complex kinetics are best represented as: (a) two components which operate throughout the exercise or (b) a delayed slow component which is consequent to the lactic acidaemia and which does not influence the early development of the O2 deficit. 2. Six healthy subjects underwent an incremental ramp test on a cycle ergometer, to the limit of tolerance, for determination of the maximum VO2 (micro VO2) and and estimation [symbol: see text] of the threshold for lactic acidaemia (theta L) non-invasively. Subjects then performed, on different days, two to four repetitions of square-wave exercise from a baseline of unloaded pedalling ('O' Watts (W)) to work rates (WR) less than theta L (90% theta L) and greater than theta L (half-way between theta L and micro VO2). Ventilatory and pulmonary gas exchange variables were determined breath-by-breath. For each subject, the VO2 transitions were averaged prior to fitting a least-squares algorithm to the on- and off-transient responses. 3. The less than theta L test resulted in a mono-exponential VO2 response, with a time constant of 31.3 and 31.5 s for the on- and off-transients, respectively. 4. The VO2 responses to the greater than theta L test were fitted to three competing models: (a) a single exponential for the entire period; (b) a double exponential for the entire period; and (c) an initial single exponential with a subsequent phase of delayed onset. Model (c) yielded a significantly lower residual mean-squares error than methods (a) and (b), with a time constant for the initial component of 40.2 s for the on-transient and 32.9 s for the off-transient and a subsequent phase of VO2 increase for the on-transient which averaged 230 ml min-1. The delta VO2/delta WR for the early kinetics of the greater than theta L test were not different from the less than theta L test (9.6 and 9.5 ml min-1 W-1, respectively). 5. These data suggest that the slow phase of the greater than theta L VO2 kinetics is a delayed-onset process. This being the case, the O2 deficit during heavy exercise, as conventionally estimated, would be overestimated.