Mechanisms underlying the upright T wave in mammalian hearts are controversial; whilst it is generally agreed that the dispersion of action potential duration (APD) throughout the ventricles lies behind the genesis of the positive T wave, the precise mechanism is still subject to debate. The variation in APD throughout the heart is owing to the heterogeneous distribution of ion channels and varying kinetics. This heterogeneity can broadly be split into three groups: transmural (TM) heterogeneity, apico-basal (AB) heterogeneity and interventricular (IV) heterogeneity (i.e. between left and right ventricles). The aim of this study was to use a biophysically detailed computer model to investigate possible contributions of TM, AB and IV heterogeneities to the positive T-wave. A family of single cell computational models was developed for rabbit ventricular myocytes in a previous study. The kinetics and conductances of the Hodgkin-Huxley equations were modified to create four distinct cell types; Purkinje fibre (PF) and endocardial, midcardial and epicardial cells of the left ventricle (LV), accounting for the transmural heterogeneity of the heart. In the present study, the LV single cell rabbit models were modified based on experimental data (Convery et al., 1998; Suto et al., 2005, 2007) to create a set of RV single cell models. Furthermore, ICa,L, IKr and IKs current densities were varied linearly to account for observed differences between apex and base (Cheng et al., 1999; Sims et al., 2008). The single cell models were incorporated into an anatomical model of the ventricles generated by a DT-MRI scan. The monodomain model was used to provide electrotonic coupling between cells, with the diffusion tensor being constructed from DT-MRI data and known conduction velocities. In simulations, the ventricles were activated by stimulating the top end of the PF network with a basic cycle length of 330 ms. The effects of each type of heterogeneity were assessed using five different configurations: 1) completely heterogeneous, 2) completely homogeneous, 3) only AB heterogeneity present, 4) only TM heterogeneity present, 5) AB and TM heterogeneity present. A limb II pseudo-ECG was calculated for each configuration, and the ECG characteristics were compared. In addition, a cross-section from the LV free wall was stimulated from the endocardial surface to simulate a ventricular wedge preparation. Our simulation data suggested that while the TM heterogeneity is sufficient to produce an upright T-wave in the ventricular wedge preparation, it is necessary to have both TM and AB heterogeneity present in order to produce a fully upright T-wave in the whole heart setting. In conclusion, this study provides mechanistic insights towards understanding the ionic basis underlying the positive T-wave in the ECG.