We summarise recent results obtained in testing some of the algorithms utilised for estimating breath-by-breath (BB) alveolar O2 transfer (VO2A) in humans. VO2A is the difference of the O2 volume transferred at the mouth minus the alveolar O2 stores changes. These are given by the alveolar volume change at constant O2 fraction (FAiO2 DeltaVAi) plus the O2 alveolar fraction change at constant volume [V(Ai-1)(FAi-F(Ai-1))O2], where V(Ai-1) is the alveolar volume at the beginning of the breath i. All these quantities can be measured BB, with the exception of V(Ai-1), which is usually set equal to the subject's functional residual capacity (FRC) (Auchincloss algorithm, AU). Alternatively, the respiratory cycle can be defined as the time elapsing between two equal O2 fractions in two subsequent breaths (Grønlund algorithm, GR). In this case, FAiO2= F(Ai-1)O2 and the term V(Ai-1)(FAi-F(Ai-1))O2 disappears. BB alveolar gas transfer was first determined at rest and during exercise at steady-state. AU and GR showed the same accuracy in estimating alveolar gas transfer; however GR turned out to be significantly more precise than AU. Secondly, the effects of using different V(Ai-1) values in estimating the time constant of alveolar O2 uptake (VO2A) kinetics at the onset of 120 W step exercise were evaluated. VO2A was calculated by using GR and by using (in AU) V(Ai-1) values ranging from 0 to FRC +0.5 l. The time constant of the phase II kinetics (tau2) of VO2A increased linearly, with V(Ai-1) ranging from 36.6 s for V(Ai-1)=0 to 46.8 s for V(Ai-1)=FRC+0.5 l, whereas tau2 amounted to 34.3 s with GR. We concluded that, when using AU in estimating VO2A during step exercise transitions, the tau2 value obtained depends on the assumed value of V(Ai-1).