To assess the prevalence of physiologic left ventricular hypertrophy and the usefulness of ECG criteria for its diagnosis, we compared ECGs and M-mode echocardiograms from 44 ultraendurance athletes and 20 similarly aged sedentary control subjects. Left ventricular mass was elevated in 25 of 44 (57%) athletes including 17 of 29 (59%) men greater than 134 gm/m2 and 8 of 15 (53%) women greater than 110 gm/m2. The sensitivity and specificity of the three ECG criteria used to diagnose left ventricular hypertrophy were: Sokolow-Lyon voltage (S-V1 + R-V5 greater than or equal to 3.5 mV), 65% and 61%; Romhilt-Estes score (greater than or equal to 4), 16% and 84%; and Cornell voltage (R-aVL + S-V3 greater than 2.8 mV in men and greater than 2.0 mV in women), 8% and 95%, respectively. Left ventricular mass, mass index, posterior wall thickness, chamber diameter, and relative wall thickness were not related to any measurement of QRS voltage. Nonvoltage ECG criteria for left ventricular hypertrophy were rare in athletes. Thus hypertrophy is a common but not universal adaptation to exercise. It is only moderately well detected by standard voltage criteria for left ventricular hypertrophy and is not reflected in nonvoltage criteria.