Analysis of the duration and terminal components of the filtered QRS complex on the signal-averaged electrocardiogram (ECG) has been widely used for the detection of late potentials. Although filtered QRS duration is strongly related to 12-lead QRS duration, the relation of performance of the signal-averaged ECG to baseline QRS duration has not been critically examined. To examine the relation of test performance of the signal-averaged ECG to 12-lead QRS duration and to test the hypothesis that the difference between filtered and baseline 12-lead QRS duration would reflect more accurately the presence of late potentials than would analysis of the filtered QRS alone, we evaluated signal-averaged and 12-lead ECGs in 144 normal subjects and in 132 patients who were examined by electrophysiologic study and of whom 45 had inducible sustained monomorphic ventricular tachycardia. The signal-averaged ECG was considered positive by standard late potential criteria when the filtered vector QRS duration was > 114 msec and either the root-mean-square voltage of the terminal 40 msec of the filtered QRS was < 20 microV or the low-amplitude signal of the terminal filtered QRS was > 38 msec. A new signal-averaged ECG criterion for the presence of late potentials was developed in the 144 normal subjects on the basis of the difference between the longest filtered QRS duration in any of the orthogonal leads and QRS duration on the baseline 12-lead ECG ("the QRS difference"), which was adjusted by regression analysis for the decreasing QRS difference found with increasing baseline QRS duration.(ABSTRACT TRUNCATED AT 250 WORDS)