Late potentials detected by the time domain signal-averaged electrocardiogram (ECG) are a well established marker for ventricular tachycardia in patients after a myocardial infarction, but the value of frequency domain analysis of the signal-averaged ECG in identifying these patients remains controversial. This study compared the results of time domain, frequency domain and spectral temporal mapping analyses of the signal-averaged ECG in 30 postinfarction patients with spontaneous sustained ventricular tachycardia and in 30 postinfarction patients without ventricular tachycardia matched for age, gender and infarct site. No patient with bundle branch block was included. Time domain signal-averaged ECG indexes were significantly different in patients with and without ventricular tachycardia (p less than 0.001). Frequency domain results were not consistently different between these groups. The values of the normality factor of spectral temporal mapping were significantly lower in patients with ventricular tachycardia (p less than 0.04). Results of the time domain signal-averaged ECG were abnormal in 22 patients with ventricular tachycardia (73%) but in only 3 control patients (10%) (p less than 0.001). Spectral temporal mapping results were abnormal in 21 patients with ventricular tachycardia (70%) compared with 12 control patients (40%) (p less than 0.04). When the optimal numeric values of dichotomy points were computed for patient stratification at different sensitivity levels, time domain analysis identified patients with ventricular tachycardia with significantly fewer false positive results than were obtained with either frequency analysis or spectral temporal mapping. It is concluded that frequency domain analysis and spectral temporal mapping of the signal-averaged ECG did not improve the identification of postinfarction patients with ventricular tachycardia and without bundle branch block.