The role of the signal‐averaged ECG was prospectively assessed in 517 patients in whom there was a suspicion for malignant ventricular arrhythmias. Patients were divided into Group I with a normal surface QRS width < 120 ms (426 patients) and Group II with a prolonged QRS duration > 120 ms (91 patients). Late potentials were present in 42 (10%) Group I patients and in 24 (26%) Group II patients. Programmed ventricular stimulation was performed for standard indications in 55 patients without late potentials and in 42 patients with late potentials, combining both groups. The sudden death or recurrent sustained ventricular tachycardia rate in follow‐up was evaluated based on the presence or absence of late potentials and whether programmed ventricular stimulation was performed. In the patients without late potentials, these rates were 4 patients (1.0%) in the no EP group and 3 patients (5.5%) in the EP group (p<.05), respectively (overall 1.6%). In the patients with late potentials, these rates were 7 patients (29%) in the no EP group and 7 patients (17%) in the EP group (p = .19), respectively (overall 21%). In addition, appropriate automatic defibrillator shocks were present in 1 patient without late potentials and in 8 patients with late potentials which were not included in the recurrent sudden death or sustained ventricular tachycardia statistics. The signal‐averaged ECG accurately defines patients at a higher risk for malignant ventricular arrhythmias regardless of unfiltered QRS duration. There is a trend toward improved survival with electrophysiology guided therapy in the late potential positive group with added incremental benefit from patients undergoing automatic defibrillator implantation.