A digital computer model is presented for the simulation of the body surface electrocardiogram (ECG) during ventricular activation and recovery. The ventricles of the heart are represented in detail by a three-dimensional array of approximately 4000 points which is subdivided into 23 regions. Excitation sequence and cellular action potential data taken from the literature are used to determine the spatial distribution of intracellular potentials at each instant of time during a simulated cardiac cycle. The moment of the single dipole representing each region is determined by summing the spatial gradient of the intracellular potential distribution throughout the region. The resulting set of 23 dipoles is then used to calculate the potentials on the surface of a bounded homogeneous volume conductor with the shape of an adult torso. Simulated isopotential surface maps during both activation and recovery are in good agreement with data for humans reported in the literature.