Objective: To determine the prevalence, impact, and stability of different subtypes of headache in a 30 year prospective follow-up study of ageneral population sample. Design: Prospective cohort study. Setting: Canton of Zurich, Switzerland. Participants: 591 people aged 19-20 from a cohort of 4547 residents of Zurich, Switzerland, interviewed seven times across 30 years of follow-up. Main outcome measures: Prevalence of headache; stability of the predominant subtype of headache over time; and age of onset, severity, impact, family history, use of healthcare services, and drugs for headache subtypes. Results: The average one year prevalences of subtypes of headache were 0.9% (female:male ratio of 2.8) for migraine with aura, 10.9% (female:male ratio of 2.2) for migraine without aura, and 11.5% (female:male ratio of 1.2) for tension-type headache. Cumulative 30 year prevalences of headache subtypes were 3.0% for migraine with aura, 36.0% for migraine without aura, and 29.3% for tension-type headache. Despite the high prevalence of migraine without aura, most cases were transient and only about 20% continued to have migraine for more than half of the follow-up period. 69% of participants with migraine and 58% of those with tension-type headache manifested the same predominant subtype over time. However, the prospective stability of the predominant headache subtypes was quite low, with substantial crossover among the subtypes and no specific ordinal pattern of progression. A gradient of severity of clinical correlates and service use was present across headache subtypes; the greatest effect was for migraine with aura followed by migraine without aura, and then tension-type headache and unclassified headaches. Conclusions: These findings highlight the importance of prospective follow-up of people with headache. The substantial longitudinal overlap among subtypes of headache shows the developmental heterogeneity of headache syndromes. Studies of the causes of headache that apply diagnostic nomenclature based on distinctions between discrete headache subtypes may not capture the true nature of headache in the general population.