It is clinically important to estimate the size of a myocardial infarction (MI) to predict patient prognosis, to determine the ability of a therapy to limit its size, and to evaluate its effect on left ventricular function. Various electrocardiographic methods have been used for these purposes but their accuracies have not been compared with each other using an identical reference population of anatomically measured infarcts. The capability of 4 electrocardiographic scoring methods (the Selvester score, the Minnesota code, the Novacode, and the Cardiac Infarction Injury Score) to estimate MI size was compared using anatomic MI size in a group of 100 deceased patients. All patients had a standard 12-lead electrocardiogram of sufficient quality to perform manual waveform measurements and without confounding factors such as ventricular hypertrophy, fascicular block, or bundle branch block. The location and size of the left ventricular infarction was measured postmortem using the anatomic method of Ideker et al. All methods' size estimates correlated best with anatomic MI size in the anterior location (r = 0.65 to 0.89). The Selvester score was superior in estimating the sizes of inferior (r = 0.70) and posterolateral (r = 0.74) infarcts. For multiple infarcts all methods performed poorly (r = 0.18 to 0.44).