The long QT syndrome is electrocardiographically characterized by a prolonged QT interval and by several other, more subtle, ST-T-U wave abnormalities, most of which have not been quantified. To determine the possible usefulness of several new electrocardiographic characteristics in identifying patients with known long QT syndrome, logistic regression models were applied to a data base of seven new, relatively independent, electrocardiographic repolarization variables. These were measured on digitized 12-lead electrocardiograms of 315 normal subjects and 37 patients with the long QT syndrome (members of well-identified long QT syndrome families, QTc greater than 0.44 second, 27% symptomatic), who ranged in age from 17 to 60 years. Electrocardiographic variables that independently differentiated (p less than 0.001) patients with long QT syndrome from normal subjects included quantitative measures of repolarization: early duration, rate, T wave symmetry, late phenomena, and heterogeneity. All selected repolarization variables except the early duration variable were essentially independent of the QTc (r2 less than 0.15), and all contributed significantly to the identification of patients with long QT syndrome. A classification model of five electrocardiographic predictor variables resulted in an estimated sensitivity (95% confidence interval) of 92.6% (81.6-100%) and an estimated specificity (95% confidence interval) of 95.8% (93.6-98.1%). This model performed significantly better than an alternative classification model that was based on the early duration variable as a single predictor variable. The symptomatic status of patients with long QT syndrome could not be predicted by any combination of the electrocardiographic variables in the investigated model.