Phrenic nerve stimulation, electrical (ES) or from cervical magnetic stimulation (CMS), allows one to assess the diaphragm contractile properties and the conduction time of the phrenic nerve (PNCT) through recording of an electromyographic response, traditionally by using surface electrodes. Because of the coactivation of extradiaphragmatic muscles, signal contamination can jeopardize the determination of surface PNCTs. To address this, we compared PNCTs with ES and CMS from surface and needle diaphragm electrodes in five subjects (10 phrenic nerves). At a modified recording site, lower and more anterior than usual (lowest accessible intercostal space, costochondral junction) with electrodes 2 cm apart, surface and needle PNCTs were similar (CMS: 6.0 +/- 0.25 ms surface vs. 6.2 +/- 0.13 ms needle, not significant). Electrodes recording the activity of the most likely sources of signal contamination, i.e., the serratus anterior and pectoralis major, showed distinct responses from that of the diaphragm, their earlier occurrence strongly arguing against contamination. With ES and CMS, apparently uncontaminated signals could be consistently recorded from surface electrodes.