It is a general belief that migraine attacks are prone to occur on days off. Only a few studies, however, have addressed this issue. The objective of this study was to investigate the periodicity of migraine with respect to weekly (circaseptan) variations. Eighty-nine females of fertile age who had participated in a previous questionnaire-based study volunteered to record in detail every migraine attack for 12 consecutive months. Eighty-four patients completed recordings for a mean of 311 days (s.d. = 95.9, range 30-365). A total of 2314 attacks were recorded. Migraine occurrence was almost equally distributed during the week, except on Sundays, when there were significantly fewer attacks (t = -4.42, d.f. = 83, P < 0.001). A Mantel-Haenszel estimate of the relative risk of having an attack on a holiday vs. another day, not Sundays included, was 0.64 (95% CI 0.49-0.85). Our study suggests that days off protect against migraine.