Provoking delayed migraine with nitroglycerin in migraine sufferers is a cumbersome model. Patients are difficult to recruit, migraine comes on late and variably and only 50-80% of patients develop an attack. A model using normal volunteers would be much more useful, but it should be validated by testing the response to drugs of known efficacy in acute migraine. Furthermore, treatment during headache rather than pretreatment is the most naturalist method. To fulfil these requirements we used continuous long-lasting infusion of glyceryl trinitrate (GTN) 0.2 microg kg-1 min-1 for 140 min and gave aspirin 1000 mg, zolmitriptan 5 mg or placebo to normal healthy volunteers. The design was double-blind, placebo-controlled three-way crossover. Our hypothesis was that these drugs would be effective in the treatment of the mild constant headache induced by long-lasting GTN infusion. The headaches did not fulfil the International Headache Society diagnostic criteria for migraine without aura. Moreover, there was no effect on headache of either zolmitriptan or aspirin. Thus our hypothesis was disproved and we conclude that our model is not valid for the testing of new acute antimigraine drugs. Our experiment suggests that headache caused by direct nitric oxide (NO) action in the continued presence of NO is very resistance to analgesics and to specific acute migraine treatments. This suggests that NO works very deep in the cascade of events associated with vascular headache, whereas tested drugs work higher in the cascade. The model suggested here should therefore be tested with other headache/migraine-provoking agents that supposedly work higher in the cascade of events. The need for human models persists, but the solution to this problem is still pending.