This study analyzes 234 patients who recovered from an initial ischemic episode. The object was to see if the duration of the first episode influenced the chance of finding a treatable lesion or the chance of a further episode. The initial episodes varied from less than 5 minutes to longer than 3 weeks. There seemed to be no fundamental difference between transient ischemic attacks (TIAs) (less than 24 hours) and strokes which recover. However, 51% of those whose initial episode lasted less than 5 minutes had a subsequent stroke compared to 28% of those with an initial episode of more than 24 hours duration. Thirty percent of the former group who had angiograms had an operable lesion against 10% in the latter group. It seems that angiography has sufficiently high yield to be warranted in all patients where the initial attack lasted less than 30 minutes. In those with longer attacks the yield from angiography was much lower and noninvasive techniques should be considered in these patients, where available, prior to consideration for angiography. Investigation should be based on the degree of functional recovery and not on the arbitrary time division which normally divides TIAs and strokes. Bruits were the most reliable clinical indicators of stenosis. However the presence of intermittent claudication, hypertension and age over 50 were all more common in those with carotid stenosis.