Normobaric hyperoxia decreases heart rate (HR) in humans and animals. This study explored the mechanisms of hyperoxic bradycardia by examining its response time, autonomic neural mediation, and reversibility in conscious dogs. Five trained mongrel dogs breathed from a mask as the inspired gas was alternated between air and O2 for multiple cycles, and continuous time series records of HR and oxyhemoglobin saturation were recorded on a digital computer and analyzed by the technique of ensemble averaging. Hyperoxia decreased HR by 9% (P < 0.001), but only gradually, requiring 5 min to reach steady state. This delay was much longer than the time required for hyperoxic respiratory depression (10-20 s), a response known to be mediated by chemoreceptor reflexes. The bradycardia was sustained for greater-than-or-equal-to 30 min. On return to normoxia, HR gradually returned toward, but failed to reach, the baseline HR, suggesting incomplete reversibility of the response. However, in control experiments without hyperoxic challenge, HR showed a slow continuous downward trend that was sufficient to account for the apparent incomplete reversibility of hyperoxic bradycardia. Hyperoxic bradycardia was unaffected by beta-adrenergic blockade but was completely prevented by muscarinic cholinergic blockade. We conclude that 1) hyperoxia-induced bradycardia in conscious dogs is mediated by efferents of the vagus nerve; 2) its afferent pathway remains unknown, but its long response time suggests mechanisms other than chemoreceptor reflexes or other known neural reflexes; and 3) it is completely reversible.