To derive a quantitative index of regional myocardial blood flow, the arterial input function of the flow tracer N-13 ammonia and the regional myocardial N-13 activity concentrations were noninvasively determined in 29 experiments in eight dogs. N-13 ammonia was administered intravenously and cross-sectional images were acquired dynamically using an ECAT III positron emission tomograph with an effective in-plane resolution of 13.46 mm full-width half-maximum. Time-activity curves were derived from the serial images by assigning regions of interest to the left ventricular myocardium and left ventricular blood pool. Tracer net extractions were estimated from the myocardial time-activity concentrations at various times after tracer injection and the integral of the arterial input function. Myocardial blood flow was altered by intravenous dipyridamole, morphine, propranolol and partial or complete occlusion of the left anterior descending coronary artery, and ranged from 9 to 860 ml/min per 100 g. Estimates of tracer net extractions were most accurate when determined from the myocardial N-13 activity concentrations at 60 s divided by the integral of the arterial input function to that time. These estimates correlated with regional myocardial blood flows determined independently by the microsphere technique by y = x (1 - 0.64(e-114/x); SEE = 22.9; r = 0.94). First pass extraction fractions of N-13 ammonia determined noninvasively with this approach declined with higher flows in a nonlinear fashion and were similar to those determined invasively by direct intracoronary N-13 ammonia injections. The findings indicate that an accurate index of regional myocardial blood flow can be obtained noninvasively by high temporal sampling of arterial and myocardial tracer activity concentrations with positron emission tomography. They also provide a basis for the in vivo application of tracer kinetic principles to derive quantitatively and noninvasively regional rates of functional processes in human myocardium.