Résumé : Body surface potential maps were recorded from 117 thoracic sites and 3 limb electrodes in 173 normal subjects older than 30 years of age and 122 patients with clinically "pure" left ventricular (LV) hypertrophy. Typical LV hypertrophy map patterns were identified at successive instants during the PQRST waveform by removing from sequential LV hypertrophy maps the corresponding normal variability range at each electrode site. The presence in individual patients of 1 or more patterns typical in time and location of LV hypertrophy allowed retrospective assignment to the LV hypertrophy group. The most consistent discriminant patterns were excessive negative voltages in the anterior torso with reciprocal excess of positive voltages in the upper right chest during the second half of the P wave, excessive negative voltages in the lower right anterior torso at mid-QRS and excessive negative voltages in the left precordium with reciprocal excess of positive voltages in the upper right chest throughout ST-T. Best classification results were achieved with ST-T features, followed by features from the P wave, the QRS waveform and the PR segment. Cumulative use of ST-T and P features yielded a specificity of 94% with a sensitivity of 88%. Little improvement was obtained by the addition of QRS and PR information. The discriminant map criteria were applied to body surface potential maps from 169 new subjects (77 normal subjects ages 20 to 30 years and 92 patients with complicated LV hypertrophy). Little modification in specificity (93%) and sensitivity (90%) was observed. The performance of commonly used standard lead criteria was also tested. The Sokoloff-Lyon criterion and the Romhilt-Estes point score produced specificities of 82 and 97%, respectively, with sensitivities of 55 and 64%. The recently described Cornell voltage criteria achieved a specificity of 83% and a sensitivity of 72%. © 1989.