Late auditory evoked potentials were recorded in four subjects during musical tasks A PDP 12 computer synchronized stimuli, which were produced by an integrated circuit, and recording with the help of a quartz time basis The content of each experiment was different modelling of an ambiguous identical acoustic stimulus In experiment 1, subjects had to model a 6-note melody according to the classic metric foot In experiment 2, segmentation of an 8note melody into 5 and 3 versus 3 and 5-tone motifs had to be performed In experiment 1 an intra-individually reliable, but inter-individually variable neurophysiological correlate was detected during the heavy tone: ( 1) positivity, ( 2) negativity, ( 3) alpha blocking and ( 4) DC shift Experiment 2 yielded an intra and inter-individually reliable positive DC shift of about 4 VIV between the two motifs Myogenic, ocular, dermal, respiratory or electrocardiographic artefacts were excluded in each case The results indicate that conclusions from evoked potentials to musical perception might be possible and that possible modelling mechanisms with subsequent undesirable influence on recordings have to be considered in any kind of evoked potential experimental design.