We examined the "vascular waterfall" hypothesis, which proposes that coronary flow is unaffected by elevations in outflow pressure until the latter reaches a critical threshold level, in 29 isolated canine hearts. In fibrillating hearts vasodilated with adenosine or carbocromen, coronary flow and the coronary pressure-flow relation were not affected by changes in great cardiac vein pressure (PGCV) below a threshold value of 11 +/- 0.9 (mean +/- SEM) mm Hg. Further elevations of PGCV reduced flow and shifted the pressure-flow relation to the right, increasing its pressure-axis intercept (Pf=0). When vasomotor tone was augmented with vasopressin, threshold PGCV increased to 25 +/- 2.7 mm Hg (p less than 0.001). Once again, the pressure-flow relation was unaffected by changes in PGCV below the threshold value and shifted to the right when this value was exceeded. The amount by which spontaneous values of Pf=0 exceeded threshold values of PGCV was greater when vasomotor tone was augmented than during vasodilation. Pf=0 continued to exceed PGCV when the latter was raised above the threshold level. Both Pf=0 and threshold values of PGCV were less during a long diastole than during ventricular fibrillation. We reached the following conclusions. 1) During changes in PGCV below a threshold value, the coronary circulation exhibits traditional waterfall behavior. 2) The threshold pressure for altering waterfall behavior is affected by vascular tone and mechanical activity. 3) Pf=0 remains above PGCV when the latter is increased above the threshold value needed to alter flow.