Left ventricular systolic pressure (LVSP) and stress in the outer layers of the left ventricular wall were directly recorded in 10 anaesthetised, open-chested dogs. Left ventricular oxygen consumption (VO2) was calculated from the difference in oxygen content in the arterial and in the coronary sinus blood and from the left ventricular coronary flow (LVCF). LVCF was measured in the shunt between the carotid artery and left coronary artery (2 experiments) or in the shunt between the coronary sinus and jugular vein. Tension-time index was calculated either as the product of the mean LVSP and time (TTI(P)), or as the product of mean systolic stress and time (TTI(sigma)). Both TTIs were changed within the broad range by means of exsanguination and blood infusion. Contractility was changed by means of Inderal or noradrenaline infusion. In all experimental conditions VO2/100 g/stroke correlated linearly (P less than 0.01) with TTI(sigma), with correlation coefficient r greater than 0.8. When TTI(P) was used, correlation coefficient r was less than 0.6 and no correlation was found in one series of experiments with noradrenaline infusion. It is concluded that TTI calculated from the directly measured wall stress is a very good correlate of the VO2, which is not the case when the 'classical' TTI is used.