Using mathematical simulation of motor unit potentials (MUPs), detected by a point and rectangular plate electrode, we have shown that the muscle tissue does not act like a low-pass frequency filter on MUPs. Depending on the electrode type and its longitudinal position, the relative weight of the terminal phases (reflecting the excitation extinction) in MUPs and thus of high frequencies in the MUP power spectrum, increase with the MU depth. Therefore, high-pass filtering or differentiating signals detected neither monopolarly nor bipolarly could eliminate the cross-talk produced by high frequency components of MUPs from deep MUs. Such methods could be effective against the main components but not against the MUP leading edge and terminal phases. To reduce the cross-talk, position of the detecting electrodes should correspond to anatomy of muscles producing the cross-talk. Monopolar electrode should be located above the ends of the muscles. Cross-talk of the muscles located beyond the muscle of interest could be higher than that produced above the end-plate of deep muscles. On the contrary, under detection by a longitudinal bipolar electrode, the cross-talk is much smaller above the end-plate region or beyond deep muscles. The cross-talk is the greatest above the ends of the deep muscles.