The automatic metabolic units calculate breath-by-breath gas exchange from the expiratory data only, applying an algorithm ('expiration-only' algorithm) that neglects the changes in the lung gas stores. These last are theoretically taken into account by a recently proposed algorithm, based on an alternative view of the respiratory cycle ('alternative respiratory cycle' algorithm). The performance of the two algorithms was investigated where changes in the lung gas stores were induced by abrupt increases in ventilation above the physiological demand. Oxygen, carbon dioxide fractions and ventilatory flow were recorded at the mouth in 15 healthy subjects during quiet breathing and during 20-s hyperventilation manoeuvres performed at 5-min intervals in resting conditions. Oxygen uptakes and carbon dioxide exhalations were calculated throughout the acquisition periods by the two algorithms. Average ventilation amounted to 6·1 ± 1·4 l min-1 during quiet breathing and increased to 41·8 ± 27·2 l min-1 during the manoeuvres (P<0·01). During quiet breathing, the two algorithms provided overlapping gas exchange data and noise. Conversely, during hyperventilation, the 'alternative respiratory cycle' algorithm provided significantly lower gas exchange data as compared to the values yielded by the 'expiration-only' algorithm. For the first breath of hyperventilation, the average values provided by the two algorithms amounted to 0·37 ± 0·34 l min-1 versus 0·96 ± 0·73 l min-1 for O2 uptake and 0·45 ± 0·36 l min-1 versus 0·80 ± 0·58 l min-1 for exhaled CO2 (P<0·001 for both). When abrupt increases in ventilation occurred, such as those arising from a deep breath, the 'alternative respiratory cycle' algorithm was able to halve the artefactual gas exchange values as compared to the 'expiration-only' approach.