To develop a reliable experimental model of vascular headache, we studied the dose-response relationship between headache and i.v. nitroglycerin (NTG) in 10 healthy subjects. NTG was infused intravenously over periods of 10 min separated by wash-out periods. Doses of 0.25, 0.50, 1.00 and 2.00 micrograms/kg/min were applied successively with one placebo infusion and wash-out period inserted randomly and double blindly. The subjects scored their headache intensity on a scale 0-10. After 1-8 weeks a retest was performed. Nine subjects developed headache already at 0.25 microgram/kg/min, whereas one had no headache at any dose. Headache severity did not increase with doses above 0.5 microgram/kg/min. This ceiling effect was reproducible. The headache was moderate, usually throbbing, bifrontal and not associated with other migrainous features. It reached maximum within 2.5-5.5 min (medians) at various doses and declined rapidly after NTG discontinuation. Wash-out periods of 10-20 min were sufficient. The reproducibility of headache intensity and character was satisfactory in the retest experiment. There were no unpleasant side effects and no visible flushing. Thus blindness was maintained. I.v. NTG is suitable as an experimental headache model. A constant infusion of 0.5 microgram/kg/min will be suitable for studies of arterial diameter, pulsations, blood flow, etc. Comparative studies of sensitivity should use the present infusion schedule but with the two highest doses substituted by 0.06 and 0.125 microgram/kg/min.