1. The effects of the non-selective cyclo-oxygenase (COX) inhibitor indomethacin and the selective COX-2 inhibitors, N-[2-(cyclohexyloxy)-4-nitrophenyl] methanesulphonamide (NS-398), 5-methanesulphonamido-6-(2,4-difluorothio-phenyl)-1-indan one (L-745,337) and 5,5-dimethyl-3-(3-fluorophenyl)-4-(4-methylsulphonyl) phenyl-2(5H)-furanone (DFU), on the protection induced by the mild irritant 20% ethanol were investigated in the rat stomach. 2. Instillation of 20% ethanol (1 ml, p.o.) effectively protected against gastric mucosal injury induced by subsequent instillation of 70% or 96% ethanol (1 ml, p.o.). 3. Oral administration of indomethacin (1.25-20 mg kg[-1]) dose-dependently counteracted the protective effect of 20% ethanol (ID50: 3.5 mg kg[-1]). 4. Likewise, NS-398 (0.1-1 mg kg[-1]), L-745,337 (0.2-2 mg kg[-1]) and DFU (0.02-0.2 mg kg[-1]) inhibited the protective effect of 20% ethanol in a dose-dependent manner with ID50 values of 0.3 mg kg(-1), 0.4 mg kg(-1) and 0.06 mg kg(-1), respectively. 5. Inhibition of mild irritant-induced protection was also found when NS-398 (1 mg kg[-1]) was administered s.c. or when 96% ethanol was used to damage the mucosa. 6. Pretreatment with 16,16-dimethyl-prostaglandin (PG)E2 at 4 ng kg(-1), a dose that did not protect against ethanol (70%)-induced mucosal damage when given alone, completely reversed the effect of the selective COX-2 inhibitors on the mild irritant-induced protection. 7. Pretreatment with dexamethasone (3 mg kg(-1), 24 and 2 h before instillation of 20% ethanol) did not affect the protective activity of the mild irritant, indicating that enzyme induction is not involved. 8. Indomethacin (20 mg kg(-1), p.o.) did not prevent the protection conferred by sodium salicylate (100 mg kg[-1]), dimercaprol (30 microg kg[-1]), iodoacetamide (50 mg kg[-1]) and lithium (20 mg kg[-1]). Likewise, the protective effect of these agents was not counteracted by NS-398 (1 mg kg(-1), p.o.). 9. Whereas indomethacin (20 mg kg(-1), p.o.) near-maximally inhibited gastric mucosal formation of PGE2, 6-keto-PGF1alpha and thromboxane (TX) B2 as well as platelet TXB2 release, the selective COX-2 inhibitors were ineffective. 10. The findings show that selective COX-2 inhibitors, although lacking in ulcerogenic activity, prevent the protection conferred by a mild irritant. Prostaglandis generated by a constitutive COX-2 could thus contribute to physiological functions involved in gastric homeostasis, although at present a non-COX-2-related mechanism underlying the effect of the selective COX-2 inhibitors tested on mild irritant-induced protection cannot be completely excluded.