While optimal diving models focus on the diver's oxygen (O(2)) stores as the predominant factor influencing diving behaviour, many vertebrate species surface from a dive before these stores are exhausted and may commence another dive well after their O(2) stores have been resaturated. This study investigates the influence of hypoxia and also hypercapnia on the dive cycle of tufted ducks, Aythya fuligula, in terms of surface duration and dive duration. The birds were trained to surface into a respirometer box after each dive to a feeding tray so that rates of O(2) uptake (VO2) and carbon dioxide output (VCO2) at the surface could be measured. Although Vco2 initially lagged behind Vo2, both respiratory gas stores were close to full adjustment after the average surface duration, indicating that they probably had a similar degree of influence on surface duration. Chemoreceptors, which are known to influence diving behaviour, detect changes in O(2) and CO(2) partial pressures in the arterial blood. Thus, the need to restore blood gas levels appears to be a strong stimulus to continue ventilation. Mean surface duration coincided with peak instantaneous respiratory exchange ratio due to predive anticipatory hyperventilation causing hypocapnia. For comparison, the relationship between surface duration and O(2) uptake in reanalysed data for two grey seals indicated that one animal tended to dive well after fully restocking its O(2) stores, while the other dived at the point of full restocking. More CO(2) is exchanged than O(2) in tufted ducks during the last few breaths before the first dive of a bout, serving to reduce CO(2) stores and suggesting that hypercapnia rather than hypoxia is more often the limiting factor on asphyxia tolerance during dives. Indeed, according to calculations of O(2) stores and O(2) consumption rates over modal diving durations, a lack of O(2) does not seem to be associated with the termination of a dive in tufted ducks. However, factors other than CO(2) are also likely to be important, and perhaps more so, such as food density and rate of food ingestion. Because some predictive success has been demonstrated for optimal diving models, they should continue to incorporate O(2) stores as a variable, but their validity is likely to be improved by also focusing on CO(2) stores.