The prevalence of late ventricular potentials (LVPs) detected by signal averaged ECG (SAECG) is variable in nonischemic heart diseases. In idiopathic dilated cardiomyopathy, the prevalence increases from about 25% to 70%-90% in cases of spontaneous sustained ventricular tachycardia (VT), is not significantly correlated with hemodynamic and Holter data, and has a good positive predictive value for induced and spontaneous sustained VT. However, its predictive value for cardiac death has not been established. In primary hypertrophic cardiomyopathy, LVPs are rare (about 10%), not correlated to hemodynamic data, enhanced in cases of spontaneous sustained VT (up to 77%), and have a good predictive value of induced VT. LVP-SAECG are frequent in arrhythmogenic right ventricular dysplasia (ARVD) (70%-80%). They can identify patients with VT and an unapparent or limited form of this disease, or ARVD with few ventricular arrhythmias. The prevalence (26%-37%) of LVPs in mitral valve prolapse is clearly higher than in normal individuals or in other valvular diseases and is enhanced in cases of spontaneous and induced VT. Its significance remains speculative. After surgical repair of tetralogy of Fallot, LVPs can identify a group of patients with higher probability of induced and spontaneous risk of VT. The usefulness and significance of LVPs in other nonischemic cardiac diseases have not to date been established. In "true" idiopathic VT, without proved structural cardiac disease, the prevalence of LVPs does not exceed that observed in normal individuals (0%-5%), but in "apparent" idiopathic VT the prevalence of LVPs rises to 20%-40%. In these latter cases more invasive techniques must be used to discover a limited form of myocardiopathy.