A new recording technique was developed to eliminate current problems on recording transcranial evoked facial muscle responses. A fork-shaped device equipped with 2 pairs of Ag/AgCl-electrodes was inserted enorally at the buccinator muscle level. Advantages offered by this method comprise clearly defined negative deflection of the compound muscle action potential, lack of relevant volume conduction from adjacent muscles, reliability of amplitude criteria, absence of interfering stimulus artifacts, easy achievement of preactivation, and noninvasive recording by surface electrodes. In 43 healthy subjects transcranial magnetic stimulation evoked contralateral responses at a mean latency and mean amplitude of 10.3 +/- 1.1 ms and 1.6 +/- 1.1 mV, respectively on the right side of the face and of 9.9 +/- 1.0 ms and 1.6 +/- 1.1 mV, on the left side of the face. Ipsilateral cortical evoked responses were observed in 29 and 25 subjects (left and right side of the face) at a mean latency and amplitude of 10.7 +/- 2.5 ms and 0.8 +/- 0.5 mV, respectively on the left side of face and of 11.9 +/- 3.2 ms and 1.1 +/- 1.2 mV, on the right side of face. No responses were obtained in 2 and 4 subjects (left and right side of the face), and could not be assessed due to simultaneous facial nerve stimulation in 12 and 14 subjects (left and right side of the face).