Prophylactic drug trials in migraine are long-lasting and expensive and require long-term toxicology information. A human migraine model would therefore be helpful in testing new drugs. Immediate headache and delayed migraine after glyceryltrinitrate (GTN) has been well characterized. We have recently shown that sodium valproate has prophylactic effect in the GTN model. Here we report our experience with propranolol in this model. Nineteen subjects with migraine without aura and 16 sex- and aged-matched healthy subjects were included in a two-centre randomized double-blind cross-over study. Fourteen migraine subjects and 14 healthy subjects completed the study and results from comparison of the 28 subjects are reported. Randomly propranolol 160 mg or placebo were each given daily for 14 days to both migraine and healthy subjects. A 20-min intravenous infusion of GTN 0.25 microg/kg per min was administered on a study day at the end of both pretreatment periods. Headache was registered for 12 h after GTN infusions. Its intensity was scored on a numerical verbal rating scale from 0 to 10. Fulfilment of International Headache Society (HIS) criteria was recorded for 24 h. Radial and superficial temporal artery diameters and blood velocity of both middle cerebral arteries were measured. All migraine subjects developed headache after GTN. No reduction of overall peak headache was found after propranolol (median 5, range 0-7) compared with placebo (median 5, range 0-10) (P = 0.441). Eight of the 14 completing migraine subject developed IHS 1.1 migraine after GTN, two subjects on both days, three subjects only after placebo, and three subjects only after propranolol. No reduction of GTN-induced migraine was found after propranolol compared with placebo (5 vs. 5, P = 1.000). All healthy subjects developed headache after GTN. No reduction of overall peak headache was found after propranolol (median 2, range 1-5) compared with placebo (median 1, range 1-7) (P = 0.315). Two subjects fulfilled IHS criteria 1.1 for migraine without aura after propranolol but not after placebo. The fulfilment was short lasting and did not require rescue medication. Headache after GTN was more pronounced in migraine subjects than in healthy subjects both with (P = 0.003) and without pretreatment with propranolol (P = 0.017). We found that 2 weeks of propranolol constricted the radial artery in healthy subjects but not in migraine subjects. GTN-induced vasodilatation abolished this difference. Mean maximum blood flow velocity in the middle cerebral artery was higher in healthy subjects than in migraine patients (P = 0.003-0.033) and unaffected by propranolol. We observed no effect of propranolol on GTN-induced headache and migraine. This could indicate that GTN induces migraine at a deeper level of the pathophysiological cascade of migraine than the prophylactic effect of propranolol. Propranolol does not constrict cerebral arteries, which therefore cannot be part of its mechanism of action in migraine.