To obtain an early prognostic diagnosis of patients with peripheral facial palsy, a magnetic stimulator (Dantec Mag 2) was used to directly stimulate the intracranial portion of the facial nerve in 15 normal subjects and 108 patients with peripheral facial palsy. In normal subjects and patients with facial palsy, compound muscle action potentials (CMAPs) of the orbicularis oris muscle elicited by transcranial magnetic stimulation were compared with CMAPs elicited by electrical stimulation at a peripheral site of the stylomastoid foramen. This technique is similar to electroneurography (ENoG) and is regularly used in our department. In normal subjects, the latency of magnetically evoked CMAPs was longer (1.0ms, SD 0.39ms) than that of CMAPs evoked by electrical stimulation. There were two categories of patients; the first group consisted of patients who visited our hospital within 2 weeks after palsy onset with a record of electrically evoked CMAPs (ENoG) and magnetically evoked CMAPs, the second group consisted of all others. The first group was then divided into four subgroups based on minimal ENoG values obtained within 2 weeks after the onset of palsy. In patients, ENoG values declined until the seventh day after palsy onset, and then plateaued. However, the amplitude ratio of magnetically evoked CMAPs between the affected side and normal side showed no tendency to deline until the seventh day after palsy onset. Thus, whether magnetically evoked CMAPs could be recorded must be discussed in relation to the prognosis of facial palsy. The patients in whom magnetically evoked CMAPs could be recorded within the seven days after the onset of palsy were classified into a group in which the minimal ENoG value was greater than 20%. These patients recovered almost 2 months after the onset of palsy, and were significantly better than the recovery rates of those patients in whom magnetically evoked CMAPs could not be recorded. The site at which the facial nerve is magnetically stimulated remains controversial. In patients with peripheral facial palsy, recovery of the stapedial reflex, blink reflex and magnetically evoked CMAPs were examined to investigate the site of magnetic stimulation. From the clinical perspective, the facial nerve is thought to be magnetically stimulated near the meatal foramen that Fisch reported the site of damage in Bell's palsy. This stimulation site was almost the same point as that calculated from the mean latency difference between magnetically evoked CMAPs and ENoG in normal controls.