Signal-averaged electrocardiograms allow the non-invasive detection of late potentials which represent locally delayed conduction in the myocardium. To validate this method, it is necessary to compare the signal-averaged data with electrograms recorded directly from the heart. However, the studies performed to date involve only a consecutive collection of the invasively and non-invasively obtained data. To obtain a more direct comparison, we examined this relation at operation by simultaneous epicardial and signal-averaged measurements. Acute infarction in animals was chosen, because the ischemic area is a zone of delayed conduction whose presence can be verified in a signal-averaged electrocardiogram. For this purpose, the left anterior descending artery, proximal of large septal and diagonal branches, was tied off in nine mongrel dogs after thoracotomy. Before infarction, a signal-averaged electrocardiogram was recorded from the body surface. At the same time, epicardial electrograms were performed using bipolar electrodes both from the supply area of the left anterior descending artery and from that of the circumflex artery. Five minutes after coronary ligation, both the epicardial measurements and the signal-averaged electrocardiogram were repeated on the open thorax. Before occlusion of the left anterior descending artery, narrow activation complexes occurred in general in the epicardial electrograms and no late potentials were recorded in any dog by the signal-averaged electrocardiogram. Five minutes after coronary ligation, fractionated and prolonged electrograms occurred in the epicardial recordings from the ischemic zone, while the activation complexes in the uninfarcted supply area of the circumflex artery remained unchanged.(ABSTRACT TRUNCATED AT 250 WORDS)