Left ventricular ejection fraction (LVEF) is a measure of ventricular function with clinical and prognostic significance and can be reliably calculated with various M-mode and two-dimensional echocardiographic formulas in selected, good quality echocardiograms. Subjective visual echocardiographic estimate of LVEF is a potentially less time consuming and more widely applicable method. In order to test its reliability, we performed a prospective blind trial in 40 consecutive patients undergoing biplane contrast ventriculography (BCV), to compare the visual estimate of LVEF during a complete echocardiogram of three independent observers with (1) cubed M-mode formula, (2) Teichholz M-mode formula, (3) length-area method from the four-chamber view, and (4) Simpson's single plane formula. BCV was the reference method. The best correlation with BCV was obtained by visual estimate [r of the three observers, respectively = 0.75; 0.84; 0.81] and M-mode measurements [r (1) = 0.8; r (2) = 0.8], but the most sophisticated methods provided the poorest estimate [r (3) = 0.54; r (4) = 0.49]. All correlation coefficients improved when good studies, defined as a definition of the endocardial surface of more than 75%, were selected (n = 23), but the differences persisted. One observer systematically estimated higher values than the other two (Friedman's test, p less than 0.01) and this interobserver variability suggests that each echocardiographer should test himself against BCV in his lab in order to apply the visual estimate method reliably.