Measurement of left atrial (LA) maximal volume (LA(max)) using two-dimensional transthoracic echocardiography (TTE) provides prognostic information in several cardiac diseases. However, the relationship between LA(max) and LA function is poorly understood and TTE is less well suited for measuring dynamic LA volume changes. Conversely, cardiac magnetic resonance imaging (CMR) and multi-slice computed tomography (MSCT) appears more appropriate for such measures. We sought to determine the relationship between LA size assessed with TTE and LA size and function assessed with CMR and MSCT. Fifty-four patients were examined 3 months post myocardial infarction with echocardiography, CMR and MSCT. Left atrial volumes and LA reservoir function were assessed by TTE. LA time-volume curves were determined and LA reservoir function (cyclic change and fractional change), passive emptying function (reservoir volume) and pump function (left atrial ejection fraction-LAEF) were derived using CMR and MSCT. Left atrial fractional change and left atrial ejection fraction (LAEF) determined with CMR and MSCT were unrelated to LA(max) enlargement by echocardiography (P = NS). There was an overall good agreement between CMR and MSCT, with a small to moderate bias in LA(max) (4.9 ± 10.4 ml), CC (3.1 ± 9.1 ml) and reservoir volume (3.4 ± 9.1 ml). TTE underestimates LA(max) with up to 32% compared with CMR and MSCT (P < 0.001). Left atrial function assessed with MSCT and CMR as LA fractional change and LAEF is not significantly related to LA(max) measured by TTE. TTE systematically underestimated LA volumes, whereas there are good agreements between MSCT and CMR for volumetric and functional properties.