Signal-averaged electrocardiography was performed in 132 subjects with various ventricular arrhythmias without clinically apparent heart disease and compared with the results obtained in a series of 45 normal subjects. The latter enabled definition of the following criteria of normality: QRS duration after averaging < 113 ms; Simson vector of the last 40 ms (RMS40) > or = 17 microV; low amplitude signal duration over 40 microV < 38 ms for a high pass bidirectional filter of 40-300 Hz. Complementary investigations (echo or angiography), performed in all cases, showed underlying abnormalities in 26 patients: 13 right ventricular dysplasias, 7 cardiomyopathies, 3 mitral valve prolapses and 3 minor congenital heart defects. These investigations were normal in the remaining 106 subjects. Of the 26 patients with cardiac disease, 15 had 3 criteria of positivity for late ventricular potentials, 6 had 2 criteria and the other 5 had no criteria of late ventricular potentials. Therefore, 81% of cases with cardiac disease had at least 2 diagnostic criteria of late ventricular potentials whereas only 4% of those without cardiac disease, and in the control group, had criteria of positivity (p < 0.001). Using these criteria, the predictive value of signal-averaged electrocardiography for the detection of underlying cardiac disease was good when 2 criteria are required for diagnosis of late ventricular potentials: sensitivity 81%, specificity 96%; predictive value of a positive test 78%, predictive value of a negative test 97%. Signal-averaged electrocardiography is therefore a good non-invasive method of diagnosing underlying cardiac disease in patients with ventricular arrhythmias without clinically apparent heart disease.(ABSTRACT TRUNCATED AT 250 WORDS)