Migraine is one of more than 200 headache disorders but stands out among these as a major cause of population ill health. In migraine epidemiology, the key variable is prevalence, but, from the perspective of public health, prevalence is uninformative without burden estimates. Here, we discuss how migraine epidemiology, from a quite recent start, has evolved into the respectable though imperfect science of today, but with the legacy that much of the large corpus of older literature is of questionable reliability. Newer studies have benefited from a universally accepted definition of migraine, while methodological developments have broadened the scope of migraine caseness, and published guidelines address important methodological issues. In the light of these developments, we question the apparent increase in migraine prevalence over time, offering explanations as to why this may be illusory. We suggest that the current best estimates are that global migraine prevalence is 14-15%, and that migraine accounts for 4.9% of global population ill health quantified in years lived with disability (YLDs). These evaluations are probably under-quantified rather than over-quantified, and YLDs are not a comprehensive measure of migraine-attributed burden. Despite these uncertainties, such high estimates of population ill health have clear implications for health policy.