A stroke registry was developed to determine the value of various clinical data in distinguishing lacunar from large vessel infarctions. Adequate localization was achieved in 98% of 246 patients with brain infarcts. These and 30 transient ischemic attack patients were followed for a median of 1082 days (range 2-1657). Follow-up data on TIA patients were invalidated by evidence of serious underreporting of TIAs in our general population. Among 212 male patients with cerebral infarcts not due to cardiogenic embolism, syphilis, migraine, vasculitis, or other unusual etiologies, 1-, 12-, and 36-month recurrence rates were 23%, 31% and 39% among patients with large vessel anterior circulation infarcts; 15%, 20% and 28% among patients with large vessel posterior circulation infarcts; and 8%, 16% and 21% among patients with lacunar anterior circulation infarcts, respectively. Six patients with posterior circulation lacunes did not experience recurrence. Comparative case fatality data were also compiled. Large vessel infarcts tended to be followed by further large vessel infarcts, usually in the same vascular distribution, whereas lacunar infarcts were not predictive of the type or location of subsequent events.