The role of the signal-averaged electrocardiogram in predicting the induction of sustained monomorphic ventricular tachycardia in high risk patients was assessed prospectively in 100 consecutive patients. Presenting diagnoses were syncope (38 patients), nonsustained ventricular tachycardia (24 patients), sustained ventricular tachycardia (25 patients) and sudden cardiac arrest (13 patients). Using programmed ventricular stimulation, 71 patients (group I) did not have and 29 patients (group II) did have inducible sustained monomorphic ventricular tachycardia. Using the signal-averaged electrocardiogram with filtering (6 dB/octave) at high pass corner frequencies of 67 and 100 Hz, the two groups were compared. The signal-averaged electrocardiogram was considered abnormal if all of the following criteria were satisfied: 1) the total filtered QRS complex duration was greater than 120 ms, 2) the duration of the terminal QRS complex of less than or equal to 20 microV was greater than or equal to 30 ms, and 3) at least one deflection (late potential) was present in this region. Differences between groups I and II in these three measures were highly significant (p less than or equal to 0.001). The sensitivity and specificity of signal averaging for predicting the induction of sustained ventricular tachycardia were 93 and 94%, respectively. Stepwise logistic regression analysis identified the signal-averaged electrocardiogram as the best predictor of induction of sustained monomorphic ventricular tachycardia, independent of left ventricular ejection fraction, presence of ventricular aneurysm, myocardial infarction and other clinical variables (chi-square = 93.2, p less than 0.0001). The signal-averaged electrocardiogram is a sensitive and specific test for the induction of sustained monomorphic ventricular tachycardia, having independent predictive value.