We have designed and are using an interactive computer-based headache interview in which patients sit in front of a terminal and answer questions posed by the computer. Data entered are summarized and printed for use by the patient's physician. Preliminary data obtained by the computer interview can be used to generate and test hypotheses. We compared patient-entered data to physicians' diagnoses for 26 patients with the diagnosis of common migraine and 20 with the diagnosis of muscle contraction headache. Contrary to common beliefs, we found that almost equal numbers in each group acknowledged (1) visual changes before some headaches, (2) headaches precipitated by alcohol, and (3) headaches that coincided with change in the weather. Furthermore, patients seemed more likely to acknowledge more than one kind of headache to the computer than to the physician.