Seasonal rhythm of migraine attacks may support a role of the suprachiasmatic nucleus of the hypothalamus in the pathophysiology of migraine. The objective of this study was to provide evidence for seasonal variation in migraine. Eighty-nine female migraineurs volunteered to record every migraine attack in detail for 12 consecutive months. Attacks associated with sleep complaints were defined as insomnia-related. By using Edwards' model for recognition and estimation of cyclic trends, time-series analysis was made. Fifty-eight patients, of which 26 had migraine without aura (MO) and 32 had migraine with aura (MA), completed the study. A total of 1840 attacks were recorded. The mean age +/- SD was 36.9 +/- 6.0. Patients with a lifetime history of MA showed marked seasonal fluctuation with more attacks in the light season compared to the dark. Time of peak was May 21. Peak/low ratio was 1.30 (95% CI: 1.08-1.55). When insomnia-related attacks (n = 312) were removed the seasonal variation became insignificant. There is a seasonal trend with more migraine attacks in the light season compared to the dark season in females with MA, but not MO, living in an arctic area. This is caused by the seasonal variation of insomnia-related attacks in patients with MA.