The intrinsic limitations of coronary arteriography to predict the physiological effects of coronary obstructions are well known. Therefore, more direct assessments of the functional significance of coronary stenoses are becoming increasingly important. Study of contrast passage by electrocardiogram-triggered digital radiography has been proposed as a way of assessing changes in myocardial perfusion. The main problems in this approach are the limited time for motionless image acquisition, the potential alteration of vascular volume between different states, and the changing flow pattern induced by contrast agents. This has led to empiric substitution of mean transit time (Tmn) by other time parameters and to representation of vascular volume by maximal contrast intensity (Dmax). To avoid these problems, intact dogs were studied during almost motionless image acquisition of 20-25 consecutive paced heart beats obtained with synchronous radiographic pulses. In this way, unequivocal and reproducible determination of Tmn was possible. Constant and maximal vascular volume was created by continuous infusion of dipyridamole, and it was proved that coronary flow in this model was not influenced by contrast injections. Flow in the circumflex artery was measured by a ring mounted and calibrated Doppler probe. In each dog, flow in the circumflex artery was varied by a balloon occluder in 12 small steps (range, 0-174 +/- 42 ml/min). Inverse appearance time (1/Tapp), Dmax, Dmax/Tapp, inverse time of maximal intensity (1/Tmax), and 1/Tmn were calculated and the relations of these parameters to measured flow were investigated. Tmn proved to be the most reliable parameter for this purpose (r = 0.97 +/- 0.02; mean +/- SD), followed by Tmax (r = 0.93 +/- 0.04). Dmax failed to represent vascular volume but, in fact, showed a moderate correlation with flow (r = 0.78 +/- 0.22), as did Tapp (r = 0.64 +/- 0.18, 0.75 +/- 0.27, and 0.59 +/- 0.26 for the three definitions of Tapp used in this study). Dmax/Tapp correlated better with flow than either component separately. Our results indicate that the mean transit time calculated by videodensitometry can be used to accurately assess changes in myocardial perfusion strictly according to the original principles of indicator dilution theory.