OBJECTIVE--To investigate the quantitative relationship, if any, between signal averaged electrocardiographic variables and echocardiographically determined left ventricular mass in hypertensive subjects. DESIGN--Cohort analytic prospective study. SETTING--University hospital. SUBJECTS--50 hypertensive subjects selected consecutively from inpatients. Patients older than 75 years, with underlying cardiac disease, with inconclusive echocardiograms with bundle branch block, or in atrial fibrillation were excluded. INTERVENTIONS--Antihypertensive therapy involving 41 patients was continued. MAIN OUTCOME MEASURES--Left ventricular mass calculated in accordance with the standards of the Penn convention. Thirteen criteria derived from combinations of signal averaged electrocardiographic X, Y, and Z Frank orthogonal leads, including voltage criteria, duration, and time-voltage integrals of the QRS complex. Four widely used standard electrocardiographic criteria for detection of left ventricular hypertrophy. RESULTS--There was no difference in the values for any of the electrocardiographic variables between patients with (n = 29) and without left ventricular hypertrophy (n = 21). The time-voltage integral of QRS in the horizontal plane was the best signal averaged variable related to left ventricular mass (r = 0.33, P = 0.019); however, the correlation with Rodstein voltage was stronger (r = 0.46, P = 0.0009). A positive correlation was also found between left ventricular indexed mass and Rodstein voltage (r = 0.43, P = 0.0019). Stepwise regression analysis revealed Rodstein voltage as the only predictor of indexed mass (P = 0.0019), and Rodstein voltage (P = 0.0022) and body weight (P = 0.011) as the only independent correlates of left ventricular mass. CONCLUSIONS--The relation between electrocardiographic variables and left ventricular mass or indexed mass is of limited value; signal averaged orthogonal leads do not improve this assessment compared with standard electrocardiographic leads.