We sought to compare the usefulness of echocardiography and magnetic resonance imaging in neonates with a borderline small left ventricle. The preoperative magnetic resonance and echocardiography studies of 20 consecutive patients (mean age 10 +/- 9 days) undergoing magnetic resonance imaging were analyzed. The diagnoses were aortic stenosis (n = 3), hypoplastic left heart complex (n = 12), and unbalanced atrioventricular septal defect (n = 5). The magnetic resonance imaging protocol included ventricular volumetry, flow measurements, and angiography. Potential left ventricular volumes, assuming an ideal geometric shape, were calculated by mathematically "unfolding" the compressed left ventricle. Left ventricular end-diastolic volume was 16.0 +/- 7.0 mL/m(2) of body surface area by echocardiography and 33.5 +/- 15.5 mL/m(2) by magnetic resonance imaging. Echocardiography consistently underestimated left ventricular volume and did not correlate with magnetic resonance. Of all echocardiographic parameters, mitral valve z-score was the best predictor of left ventricular end-diastolic volume by magnetic resonance (r = 0.77; P = .02). The average potential volume increase was 8.8% for aortic stenosis, 35.0% for atrioventricular septal defect and 23.0% for hypoplastic left heart complex patients. Aortic valve diameter did not correlate with flow volume in the ascending aorta. Sixteen (80%) of 20 patients underwent biventricular repair, without early mortality. Of these, only 5 (31.3%) had a preoperative left ventricular end-diastolic volume of more than 20 mL/m(2) by echocardiography. Magnetic resonance imaging is feasible in neonates with borderline left ventricular hypoplasia. Echocardiography does not accurately measure left ventricular hypoplasia in these patients and may unfairly preclude some patients from a biventricular repair in whom magnetic resonance is reassuring.