In order to compare different criteria for the definition of late potentials in patients after myocardial infarction, three signal averaged ECG variables, duration of the signal-averaged QRS complex (QRS), root-mean-square voltage of the terminal 40 ms (RMS-40), and the duration of low amplitude signals less than 40 microV (LASD-40), were assessed in 332 survivors of acute myocardial infarction who were followed-up for at least 6 months, during which 12 patients died suddenly and 14 suffered symptomatic sustained ventricular tachycardia. The associations of the three variables with arrhythmic events were analysed in the total population, in infarct site and age-specific subgroups. The sensitivity and specificity for the prediction of arrhythmic events was computed (for all dichotomy points) and compared with nine published criteria for late potentials based on the same three variables. Analysis showed that (a) the total signal averaged QRS duration was a better predictor of arrhythmic events than the other two variables, (b) for arrhythmic events in cases of anterior infarctions, higher RMS-40 dichotomy limits and lower QRS and LASD-40 dichotomy limits were needed for cases of inferior infarction, (c) a multivariate stratification of arrhythmic events based on all three variables performed better in the anterior infarction population than in the inferior infarction population, (d) the strategy defining late potentials, which requires that two variables reach critical values, is better than the strategies that require that any one or all three variables reach critical values, (e) all the definitions of late potentials performed differently in the populations with anterior as compared to inferior infarctions; to identify groups at similar risk of arrhythmic events, different criteria defining late potentials should be used in these subpopulations.