Signal-averaged electrocardiography, resting radionuclide ventriculography and Holter monitoring were performed prior to hospital discharge, to assess their value in predicting recurrent cardiac events in 210 survivors of acute myocardial infarction. In addition, 153 of these patients also underwent exercise radionuclide ventriculographic assessment. During median follow-up of 14 months (6-24 months), there were 16 cardiac deaths, 15 patients had recurrent infarction and 7 patients represented with symptomatic ventricular tachycardia. Cox regression analysis identified independent predictors of 'ischemic events' (death or re-infarction) as a previous history of infarction (p = 0.01), Killip class III-IV (p = 0.03) and an abnormal exercise radionuclide study (p = 0.04); and predictors of 'arrhythmic events' (sustained ventricular tachycardia or sudden death) as an abnormal signal-averaged electrocardiograph (p = 0.01) and left ventricular ejection fraction less than 40% (p = 0.03). Patients with an abnormal signal-averaged electrocardiograph and reduced left ventricular ejection fraction had a 34% incidence of arrhythmic events during the first 6 months compared with a 4% incidence among patients without late potentials. In those patients who underwent exercise testing and signal averaging, 85% of total cardiac events and all cardiac deaths were predicted by an abnormality of either noninvasive test. In addition, exercise testing and signal-averaged ECG were independent predictors of outcome. Hence, using a combination of noninvasive tests, patients can be stratified according to the risk of recurrent life-threatening cardiac events after myocardial infarction; such patients may be suitable for intensive investigation and considered for trials involving active intervention.