An experimental canine model of pericardial effusion was designed to validate previous clinical nuclear magnetic resonance imaging (NMR) studies. Saline (n = 7), serum (n = 4), blood (10% hematocrit [n = 5]; 20% hematocrit [n = 5]), and lipid (n = 4) effusions were chosen to resemble: (1) transudative/exudative, (2) nonhemorrhagic/hemorrhagic, and (3) chylous effusions, respectively. There was a linear correlation between the infused volume and the pericardial/epicardial distance measurements on the nuclear magnetic resonance images. Hemorrhagic and nonhemorrhagic exudative effusions were distinguished from transudative effusions by the low signal intensity of transudative effusions images obtained at a TR (repetition time) of 400 and 800 msec. Nonhemorrhagic effusions had significantly lower effusion-to-myocardial signal intensity ratio at TR of 400 msec than did hemorrhagic effusions. Differences in hematocrit were not appreciated qualitatively or quantitatively. Compared with other effusion types, only chylous effusions were hyperintense to myocardium at a TR of 400 msec. Chylous effusions were further uniquely characterized by a decreasing effusion-to-myocardial signal intensity ratio with increasing TR. These experimental findings corroborate the findings of earlier clinical reports and suggest that NMR can provide important assistance in the evaluation of pericardial effusions.