We assessed the reliability of the diagnosis of migraine with aura (MA) and migraine without aura (MO) based on the third edition of the deCODE Migraine Questionnaire (DMQ3) using a physician-conducted interview as an empirical index of validity. Amongst Danish migraine families recruited from specialist practice we selected 200 cases diagnosed according to the International Classification of Headache Disorders 2nd Edition in a validated physician-conducted telephone interview: 50 patients with exclusively MA, 50 with both MA and MO, 50 with exclusively MO and 50 controls. A written copy of the DMQ3 was mailed to the participant. The DMQ3-based diagnosis was compared with the interview-based diagnosis. Overall, the DMQ3 diagnosed migraine (MA, MO or both) with a sensitivity of 99% (109/110), a specificity of 86% (32/37) and a kappa statistic of 0.89. The most reliable subtype of migraine was MA (with or without co-occurring attacks of MO) which was diagnosed with a sensitivity of 92% (71/77), a specificity of 93% (65/70) and a kappa statistic of 0.85. Exclusively MO was diagnosed with a sensitivity of 91% (30/33), a specificity of 93% (106/114) and a kappa statistic of 0.80. Weakest was the diagnosis of both MO and MA which was diagnosed with a sensitivity of 63% (24/38), a specificity of 92% (100/109) and a kappa statistic of 0.57. In conclusion, the DMQ3 is a valid tool for diagnosing patients with migraine for genetic studies.