Résumé : The relationship between left ventricular ejection time, and physiological or biometrical data, is submitted to multivariate statistical analysis. The methodology is intended to attain 3 objectives: to simulate a controlled condition experiment in clinical research, by using the statistical method of multiple regression; to obtain a wide range of conclusions, by selecting a sample of 200 subjects whose biological data cover a wide field; to check the prospective value of results, by applying the formula derived from one group of 100 subjects to another independent group of 100 subjects. The prediction formula for ejection time takes into account stroke volume, cardiac cycle duration, body height, sex, age, pulse pressure, blood capacity in oxygen, and the presence or absence of digitalis treatment or atrial fibrillation. The first aim of the formula is to better evaluate the normal relationship between the variables considered. Secondarily, it leads to a proposal for estimating mechanical performance of the left ventricle.