Non-invasive determination of left-to-right shunts at the atrial level was performed by a new procedure using first-pass radionuclide measurement of cardiac output of the right and left ventricle. In 23 patients with coronary artery disease without shunt the mean difference between the cardiac outputs of the right and left ventricle was 0.04 l.min-1, the limits of agreement -0.80 to 0.88 l.min-1 and the 95% confidence interval for the bias -0.14 to 0.22 l.min-1. Right-sided cardiac catheterization was performed to assess the severity of the arterio-venous shunt by oximetry in 18 adult patients with atrial septal defects. The range of pulmonary-to-systemic flow ratios was 1.4 to 5.0 (mean 2.7) by oximetry and 1.3 to 4.4 (mean 2.5) by the radionuclide technique. The mean difference between flow ratios measured by the two methods was 0.25, the limits of agreement, -0.85 to 1.35 and the 95% confidence intervals for the bias -0.02 to 0.52. In two patients the difference in flow ratios was more than 1. In one of these the oximetric value was more than 2, and the radionuclide value less than 2. The limits of agreement between flow ratios determined by repeated measurements were -1.05 to 0.97, and by two independent observers -0.75 to 0.77. The present findings suggest that first-pass radionuclide determination of left-to-right shunts through atrial septal defects is both reproducible and accurate when compared to the oximetric technique.