We studied the effects of sleep on auditory evoked magnetic fields following pure tone stimulation applied to the right ear of 10 healthy normal volunteers to investigate the changes in the processing of auditory perception in the primary auditory cortex. Dual 37-channel biomagnetometers were used to record auditory evoked magnetic fields over the bilateral temporal lobes in response to presented tones. Auditory evoked magnetic fields were compared for three stimulus frequencies (250, 1000 and 4000 Hz) and three sleep stages (awake state, sleep stages 1 and 2). Four main components, M50, M100, M150 and M200, were identified with latencies of approximately 50, 100, 150 and 200 ms, respectively. The latency of each component had a tendency to be prolonged with the depth of sleep stage in all frequencies. The amplitude ratios of the early-latency components (M50 and M100) showed a tendency of reduction compared with the same components in the awake state. By contrast, the amplitude ratios of the long-latency components (M150 and M200) were significantly enhanced with an increase in the sleep stage compared with the same components in the awake state. The equivalent current dipoles of all components in all conditions were detected at the superior temporal cortex (the primary auditory cortex). As for the changes in the equivalent current dipole location of each component, the equivalent current dipole was detected in the more posterior and medial region in responses to the high-frequency tone (1000 and 4000 Hz) compared with those to 250 Hz tone stimulation. Although the equivalent current dipoles of the early-latency components (M50 and M100) were in regions more anterior and superior compared to those in the awake state, there was no consistent tendency of changes in equivalent current dipole locations between each sleep stage in the late-latency components (M150 and M200). These findings are probably due to the difference in generating mechanisms between the early- and late-latency components.