In order to test the hypothesis that glycogen sparing observed early during exercise following caffeine ingestion was a consequence of tighter metabolic control reflected in faster VO2 kinetics, we examined the effect of caffeine ingestion on oxygen uptake (VO2), carbon dioxide production (VCO2) and expiratory ventilation (VE) kinetics at the onset of both moderate (MOD) and heavy (HVY) intensity exercise. Male subjects (n = 10) were assigned to either a MOD (50% VO2,max, n = 5) or HVY (80% VO2,max, n = 5) exercise condition. Constant-load cycle ergometer exercise was performed as a step function from loadless cycling 1 h after ingestion of either dextrose (placebo, PLAC) or caffeine (CAFF; 6 mg (kg body mass)-1). Alveolar gas exchange was measured breath-by-breath. A 2- or 3-component exponential model, fitted through the entire exercise transient, was used to analyse gas exchange and ventilatory data for the determination of total lag time (TLT: the time taken to attain 63% of the total exponential increase). Caffeine had no effect on TLT for VO2 kinetics at either exercise intensity (MOD: 36 +/- 14 s (PLAC) and 41 +/- 10 s (CAFF); HVY: 99 +/- 30 s (PLAC) and 103 +/- 26 (CAFF) (mean +/- S.D.)). TLT for VE was increased with caffeine at both exercise intensities (MOD: 50 +/- 20 s (PLAC) and 59 +/- 21 s (CAFF); HVY: 168 +/- 35 s (PLAC) and 203 +/- 48 s (CAFF)) and for VCO2 during MOD only (MOD: 47 +/- 14 s (PLAC) and 53 +/- 17 s (CAFF); HVY: 65 +/- 13 s (PLAC) and 69 +/- 17 s (CAFF)). Contrary to our hypothesis, the metabolic effects of caffeine did not alter the on-transient VO2 kinetics in moderate or heavy exercise. VCO2 kinetics were slowed by a reduction in CO2 stores reflected in pre-exercise and exercise endtidal CO2 pressure (PET,CO2) and plasma PCO2 which, we propose, contributed to slowed VE kinetics.