ECG criteria for left ventricular hypertrophy (LVH) were mostly validated using left ventricular mass (LVM) as measured by M-mode echocardiography. LVM as measured by cardiac MRI has been demonstrated to be much more accurate and reproducible. We reevaluated the sensitivity and specificity of 4 ECG criteria of LVH against LVM as measured by cardiac MRI. Patients with systemic hypertension (n=288) and 60 normal volunteers had their LVM measured using a 1.5-Tesla MRI system. A 12-lead ECG was recorded, and 4 ECG criteria were evaluated: Sokolow-Lyon voltage, Cornell voltage, Cornell product, and Sokolow-Lyon product. Based on a cardiac MRI normal range, 39.9% of the hypertensive males and 36.7% of the hypertensive females had elevated LVM index. At a specificity of 95%, the Sokolow-Lyon product criterion had the highest sensitivity in females (26.2%), the Cornell criterion had the highest sensitivity in males (26.2%), and the Cornell product criteria had a relatively high sensitivity in both males and females (25.0% and 23.8%). Receiver operating characteristic curves showed the Cornell and Cornell product criteria to be superior for males whereas the Sokolow-Lyon product criterion was superior for females. Comparing the mean LVM index values of the subjects who were ECG LVH positive to the normal volunteers indicated that the ECG LVH criteria detect individuals with an LVM index substantially above the normal range. We have redefined the partition values for 4 different ECG LVH criteria, according to gender, and found that they detect subjects with markedly elevated LVM index.