Of the various waveforms in the electrocardiogram (ECG), the U wave has been the most elusive. After the first description of a U wave by Einthoven several hypotheses were put forward as to its origin. Three of these are frequently quoted, ie: 1) the repolarization of the Purkinje fibres; 2) the prolonged repolarization of the M-cells in the midmyocardium; and 3) after-potentials, possibly caused by mechanical forces in the ventricular wall. However, none of these hypotheses has gained general acceptance. We present a simple multilayered digital model of the myocardium, which explains the formation of the U wave on the basis of known electrophysiological processes responsible for the electrical sources in the myocardium, and of the physical laws, formulated in the lead vector concept, which link the potentials in or on the body to these sources. A realistic action potential (AP) is assigned to each layer. The timing of the APs is such that a normal ventricular wall activation is simulated. The differences in APs between adjacent layers create current sources Di that contribute to the potential course at an arbitrary observation point P through the heart vector-lead vector relationship. Assuming a homogeneous infinite medium, without changing the AP shapes or durations and without introducing after potentials, different realistically shaped T and U waves are simulated. Their amplitudes and configurations are dependent on the value of L, viz. the relative distance of the observation point to the myocardium. The gradual and varying transition from T wave into the U brings into question the traditional view that the end of T wave represents the end of the myocardial repolarization: T and U together must be considered as one repolarization complex. The traditional concept of QT prolongation would then need revision.