Abstract Recent studies suggest that intermittent and prolonged normobaric hyperoxia (HO) results in brain ischemic tolerance (BIT), reducing ischemic brain injury. We have attempted to determine the time course of HO-induced BIT, and to explore the putative roles of tumor necrosis factor-α (TNF-α) converting enzyme (TACE), TNF-α, and nuclear factor-κB (NF-κB) activation in mediating this effect. Two core experimental protocols were applied to rats (experiments 1 [E1] and 2 [E2] respectively). E1 rodents comprised six subgroups, breathing room air (RA; O 2=21%), or 95% oxygen (HO) for 4, 8, 16 h (4RA, 8RA, 16RA and 4HO, 8HO, 16HO respectively). E2 rodents were divided into subgroups, exposed to 95% inspired HO for 4 h/day for six consecutive days (intermittent hyperoxia, InHO) or for 24 continuous hours (prolonged hyperoxia, PrHO). Each of these had a control group exposed to 21% oxygen in the same chamber. Twenty-four hours after pretreatment, each group was randomly divided to receive 60 min right middle cerebral artery occlusion (MCAO-operated), sham-operation (without MCAO), or no operation (intact). After 24 h reperfusion, neurologic deficit score (NDS), brain water content, Evans Blue extravasation (as a marker of blood–brain barrier permeability), TACE expression, serum TNF-α, and phosphor- κBα levels were assessed in all animals, and infarct volume in the MCAO-operated subgroups. E1: Compared with the control (RA) group, infarct volume was reduced by 58.6% and 64.4% in 16 h and 24 h respectively. NDS and Evans Blue extravasation was also reduced in 16 h and 24 h. There was no statistical difference among 4 h and 8 h. E2: Preconditioning with prolonged and intermittent HO decreased NDS, infarct volume and upregulated TACE and increased phosphor-κBα and serum TNF-α level significantly. Although further studies are needed to clarify the mechanisms of brain ischemic tolerance, InHO and PrHO may partly exert their effects via triggering TACE/TNF-α/NF-κB.