The value of time-domain and spectral turbulence analyses of the signal-averaged electrocardiogram (SAECG) for predicting induction of sustained monomorphic ventricular tachycardia (VT) was prospectively investigated in 70 patients with idiopathic dilated cardiomyopathy. Sustained VT was induced in 9 patients (13%). The prevalence of abnormal time-domain and spectral analyses was 16 and 37%, respectively. The total predictive accuracy of time-domain and spectral analyses for VT induction was 86 and 67%, respectively (p < 0.01). The predictive accuracy of time-domain and spectral analysis was similar in patients without an intraventricular conduction defect (94 and 84%, respectively). However, the predictive accuracy of time-domain was higher than that of spectral analysis in patients with an intraventricular conduction defect (65 vs 25%; p < 0.05). The poor concordance between spectral analysis and programmed stimulation results was mainly due to the high number of false-positive recordings in the presence of an intraventricular conduction defect (9 of 20 cases). With the use of stepwise discriminant function analysis, an abnormal time-domain SAECG was the only variable predicting the induction of sustained VT (p < 0.0003). In dilated cardiomyopathy, an abnormal time-domain SAECG and induced sustained VT are rare, both time-domain signal-averaged electrocardiography and spectral analysis have a high predictive accuracy for VT induction in patients without an intraventricular conduction defect, and spectral analysis does not improve VT prediction in those with a conduction defect.