Training induces changes in cardiac structure and function which improves cardiac output (CO) and oxygen delivery during exercise. It is unclear whether it is cardiac structure or function which is of greatest importance in determining maximal oxygen consumption (VO(2max)). In 55 subjects (15 non-athletes, 32 amateur and 8 elite athletes), left and right ventricular (LV and RV) volumes and mass were assessed by magnetic resonance imaging (CMR). Comprehensive traditional and novel echocardiographic measures included colour-coded Doppler echocardiography to assess myocardial velocities, strain and strain rate at rest and maximal exercise in both ventricles. Measures of cardiac size and function were assessed as univariate and multivariate predictors of VO(2max). LV and RV mass correlated strongly with VO(2max) (r = 0.79 and r = 0.65, respectively, p < 0.0001), as did LV and RV end-diastolic volumes (r = 0.68 and r = 0.75, p < 0.0001) and heart rate reserve (r = 0.60, p < 0.0001). Measures of myocardial function were not predictive of VO(2max) with the exception of RV diastolic velocities (r = 0.32 and r = 0.36 for rest and exercise, respectively, p < 0.05). On multivariate analysis, only RV end-diastolic volume, LV mass and heart rate reserve were independent predictors (beta = 0.28, 0.45 and 0.27 respectively, p < 0.0001) and together explained 73% of the variance in VO(2max). Measures of cardiac morphology are strongly associated with VO(2max) in healthy adults and well-trained athletes. A combination of ventricular volume, mass and heart rate reserve explains much of the variance in VO(2max), whilst measures of myocardial function do not further strengthen predictive models.