Coronary angiography permits identification of stenotic lesions but underestimates their severity and does not provide information regarding their physiologic significance. Evaluation of coronary flow reserve by means of selective coronary artery Doppler flow catheters or quantitative arteriography has been proposed to obtain this information. However, these techniques may not accurately reflect transmural gradients in flow. We evaluated the relationship between flow reserve defined with an epicardial Doppler flow probe and the transmural gradient of flow measured with radiolabeled microspheres in 21 dogs with graded stenoses and correlated results with coronary artery geometry measured morphometrically. Four groups of dogs were studied. In five control dogs without stenosis, reactive hyperemia after 20 seconds of complete coronary occlusion was 4.5 ± 1.5 (mean ± SD) times resting flow with an endocardial/epicardial flow ratio at peak flow of 1.0 ± 0.2. When reactive hyperemia was blunted (without affecting resting flow) by 50% (n = 6), 75% (n = 5), or was abolished completely (n = 5) by coronary stenosis, the endocardial/epicardial flow ratio at peak flow was 1.0 ± 0.3, 0.7 ± 0.2, and 0.5 ± 0.1, respectively. Cross-sectional area of the stenosed segment was reduced by 85.6 ± 3.5%, 91.1 ± 2.2%, and 92.8 ± 4.3% in these groups, respectively. Thus in dogs with stenoses exceeding 86% of the cross-sectional area, endocardial flow reserve is compromised disproportionately compared with epicardial flow reserve, suggesting that clinical measurements of coronary flow reserve may underestimate the physiologic significance of coronary stenoses.