The electric heart activity can be localised from body surface mapping data with computer algorithms. At higher heart rates the T and P waves merge. Thus, the offset can not be subtracted in the TP segment. We investigated 28 healthy volunteers with signal averaged 31-lead magnetocardiography. The offset of the baseline was determined in the TP-segment and in the PR-segment, respectively. The electrical heart activity was localised in the initial 30 ms of the QRS complex (Q), at the QRS maximum (R), and at the T wave maximum (T). The volume currents were considered by using a boundary element model with the compartments lungs and torso. The 3D positions of the dipoles, the dipole orientations, and the dipole strengths were calculated using the data preprocessed with two different offset correction intervals. The offsets of the TP and PR segments significantly differed one from another. The average deviations of the dipole localisation were within a few centimetres (Q: 20 +/- 31 mm, R: 6 +/- 13 mm, T: 14 +/- 30 mm). However, in a small number of subjects (Q: n = 5, R: n = 2, T: n = 5) we observed a deviation of more than 30 mm. These deviations were not linearly correlated to the differences in the baseline offsets. High resolution recordings continuously detect heart activity in the PR segment. The correction of the baseline in the PR segment instead of the TP segment may introduce artefacts in the source localisation and therefore should be avoided.