The utility of the 12 lead electrocardiogram (ECG) in identifying the site of origin of sustained ventricular tachycardia in patients with previous myocardial infarction was studied. A new mapping grid, based on biplanar fluoroscopic imaging of the heart, was utilized for the definition of left ventricular endocardial sites. On the basis of QRS configurations resulting from left ventricular endocardial pacing at disparate sites in 22 patients (Group I), ECG features that were specific for particular sites were identified and used to construct an algorithm. Apical and basal sites were differentiated by the QRS configuration in leads V4 and aVR, anterior and inferior sites by that in leads II, III and V6 and septal and lateral sites were differentiated using leads I, aVL and V1. The algorithm was used to predict the site of earliest endocardial activation during 44 episodes of sustained ventricular tachycardia in a second group of 42 patients (Group II) in a blinded fashion. Anterior sites were correctly predicted in 83% of cases, inferior sites in 84%, septal sites in 90% and lateral sites in 82% of cases. Apical and basal sites were each correctly predicted in 70% of cases, whereas intermediate sites were less well predicted (29 to 55%) on the basis of QRS configuration. Precise localization of the site of origin of ventricular tachycardia (in all three planes) was achieved in 17 cases (39%), and in 16 cases (36%) the site of origin was immediately adjacent to the predicted site. Prediction of the site of origin of ventricular tachycardia from the 12 lead ECG may serve as a useful, time-saving adjunct to, but not a substitute for, activation sequence mapping during ventricular tachycardia.