Positron emission tomography (PET) was applied to the measurement of myocardial perfusion using the perfusion tracer 13N-labeled ammonia. 13N ammonia was delivered intravenously to 13 healthy volunteers both at rest and during supine bicycle exercise. Dynamic PET imaging was obtained in three cross-sectional planes for 10 minutes commencing with each injection. The left ventricle was divided into eight sectors, and a small region of interest was assigned to the left ventricular blood pool to obtain the arterial input function. The net extraction of 13N ammonia was obtained for each sector by dividing the tissue 13N concentration at 10 minutes by the integral of the input function from the time of injection to 10 minutes. With this approach for calculating net extractions, rest and exercise net extractions were not significantly different from each other. To obviate possible overestimation of the true 13N ammonia input function by contamination by 13N-labeled compounds other than 13N ammonia or by spillover from myocardium into blood pool, the net extractions were calculated using only the first 90 seconds of the blood and tissue time-activity curves. This approach for calculating net extractions yielded significant differences between rest and exercise, with an average ratio of exercise to rest of 1.38 +/- 0.34. Nonetheless, the increase was less than predicted from the average 2.7-2.8-fold increase in double product at peak exercise or the 1.7-fold increase in double product at 1 minute after exercise. However, when the first 90 seconds of dynamic data were fit with a two compartment tracer kinetic model, average perfusion rates of 0.75 +/- 0.43 ml/min/g at rest and 1.50 +/- 0.74 ml/min/g with exercise were obtained. This average increase in perfusion of 2.2-fold corresponded to similar average increases in double product. Thus, the noninvasive technique of PET imaging with 13N ammonia shows promise for future applications in determining absolute flows in patients with coronary artery disease.