1. The mechanisms underlying the exercise hyperpnoea have been difficult to define. Recently it has been suggested that exercise ventilation (VE) changes in proportion to changes in arterial potassium concentration ([K+]a). Similar VE and [K+]a time courses following work rate changes have been cited as supporting evidence. This study compared [K+]a and VE dynamics during moderate exercise in man. 2. We observed VE and gas exchange responses in five healthy men to sinusoidal work rate variation between 25 and approximately 105 W. Tests of approximately 30 min duration were performed at sinusoidal periods of 9, 6 and 3 min and in the steady state. In each test, during two or three sine periods, arterial blood was sampled (24 per test) and analysed for [K+] and blood gases. Response amplitude and phase (relative to work rate) were determined for each variable. 3. [K+]a fluctuated in response to sinusoidal work rate forcing with mean-to-peak amplitude averaging 0.15 mmol 1(-1). However, among tests, VE amplitude and phase were not highly correlated with [K+]a (r = 0.36 and 0.67, respectively). Further, average [K+]a amplitude in the 9 and 6 min sinusoidal studies tended to exceed the steady-state amplitude, while average VE amplitude fell progressively with increasing forcing frequency. The dissimilar dynamics of [K+]a and VE seem inconsistent with a major role for [K+]a as a proportional controller of ventilation during non-steady state moderate exercise in man. 4. Among tests, VE and CO2 output (VCO2) amplitude and phase were closely correlated (r = 0.87 and 0.94, respectively). Further, arterial CO2 pressure (Pa,CO2) and arterial pH(pHa) did not fluctuate significantly in ten of twenty and thirteen of twenty studies, respectively. In tests where sinusoidal fluctuation was detected, amplitude averaged 1.1 mmHg and 0.008 units, respectively. Thus VE demonstrated a close dynamic coupling to CO2 output, with consequent tight regulation of Pa,CO2 and pHa.