The proposal that diastolic coronary flow is regulated by an intramyocardial "back-pressure" that substantially exceeds coronary venous and ventricular diastolic pressures has been examined in an open-chest canine preparation in which instantaneous left circumflex pressure and flow could be followed to cessation of inflow during prolonged diastoles. Despite correlation coefficients consistently >0.90, pressure-flow data during individual diastoles were concave to the flow axis before and during pharmacologically induced maximum coronary vasodilation. Data were better fitted (P < 0.01) by second-order equations than by linear equations in >90% of cases. Second-order pressure-axis intercepts (P(f=0))(1) averaged 29+/-7 (SD) mm Hg before vasodilation and 15+/-2 mm Hg during vasodilation; left and right atrial pressures were always substantially lower (8+/-3 and 5+/-2 mm Hg before vasodilation and 8+/-2 and 4+/-1 mm Hg during dilation). Values of P(f=0) before vasodilation varied directly with levels of coronary inflow pressure. A modification of the experimental preparation in which diastolic circumflex pressure could be kept constant was used to evaluate the suggestion that P(f=0) measured during long diastoles are misleadingly high because of capacitive effects within the coronary circulation as inflow pressure decreases. Decreases in P(f=0) attributable to capacitive effects averaged only 5.9+/-3.0 mm Hg before vasodilation and were smaller during dilation. We conclude that P(f=0) is a quantitatively important determinant of coronary driving pressure and flow, resulting from both factors related to, and independent of, vasomotor tone. Adjustments of flow during changing physiological situations may involve significant changes in P(f=0) as well as in coronary resistance.