OBJECTIVE To investigate the effectiveness and limitations of magnetic resonance imaging in defining cardiac anatomy in patients with double inlet ventricle or atrioventricular valve atresia. METHODS Magnetic resonance images were reviewed retrospectively without reference to other morphological data. METHODS A tertiary referral centre for paediatric cardiology. METHODS 18 patients (aged 8 days to 27 years) with a suspected univentricular atrioventricular connection. METHODS Imaging by a 1.5 T whole body magnetic resonance system with imaging planes adjusted to individual patient anatomy to best define the cardiac morphology. A complete sequential diagnosis obtained from an independent interpretation of the images was compared with the diagnosis obtained from cross sectional echocardiography and angiocardiography. RESULTS There was substantial accord between the diagnosis from magnetic resonance alone and that from other methods. In the six instances where there was not accord the magnetic resonance diagnosis was considered to be correct in two cases and incorrect in three cases. In the remaining case no consensus could be reached. In eight patients magnetic resonance imaging provided anatomical information additional to that from other methods. The strengths of magnetic resonance were in imaging the pulmonary arteries and their abnormalities and identifying juxtaposed atrial appendanges but there were some deficiencies in identifying Blalock-Taussig shunts. CONCLUSIONS Magnetic resonance imaging provided detailed information about all aspects of cardiac morphology in patients with a suspected diagnosis of univentricular atrioventricular connection. Often it provided additional information to echocardiography. Its use in selected patients should give valuable complementary information.