Cerebral blood flow is thought to decrease during the preheadache phase of migraine and increase during the headache phase. Most investigations of the pathophysiology of migraine have been concerned with the factors that trigger the preheadache phase. The present hypothesis proposes that, regardless of the triggering factors associated with vasospasm, this will be followed by a common event--namely, reactive hyperaemia due to hypoxia. It is suggested that adenosine triphosphate (ATP), perhaps released from "purinergic" nerves, and its breakdown products adenosine monophosphate (AMP) and adenosine are strong contenders for agents mediating this vasodilatation. ATP is a potent dilator of cerebral vessels and its breakdown products are also dilators. High concentrations of AMP and adenosine have been collected in cerebrospinal fluid during vasodilatation following ischaemia or hypoxia. The presence of ATP and its breakdown products could also explain pain in migraine. These substances stimulate primary afferent nerve terminals in the skin and produce pain in human skin blisters; and nerve profiles that resemble afferent terminals in their ultrastructure have been described in the adventitia of cerebral arteries and in the subarachnoid meshwork. The asymmetrical nature of migraine headaches, the changes in platelet aggregation, and the responses to several therapeutic procedures are consistent with this hypothesis. Some analogues of ATP might have beneficial effects on migraine headache.