None of the current electrical audiometric procedures, alone or in combination, has yet achieved the precision of conventional audiometric testing that is used to assess hearing in verbally capable children and adults. The reason for this, in part, lies in the use of stimuli which have a wide frequency content. We have measured scalp potentials which follow the envelopes of sinusoidally amplitude-modulated tones: a frequency-specific stimulus. In normal subjects such amplitude-modulation following responses (AMFRs) appear to be generated by two sources. One source has a latency of about 30 ms, generates large responses and is only observed at modulations below 55 Hz, while the other source has a latency of 7-9 ms, generates smaller responses, and is only observed at modulations from 100-350 Hz. The latencies of these two sources are consistent with origins in the cortex and midbrain, respectively. We examined AMFRs to low frequency (50 Hz) modulations as a possible audiometric tool. In normal subjects, the amplitude of the AMFR increased as a function of intensity, decreased as a function of carrier frequency, and could be evoked across the whole audiometric range (250-8000 Hz). In hearing-impaired subjects, the AMFR amplitudes as a function of carrier frequency accurately reflected the pattern of hearing loss on a frequency-by-frequency basis. In most subjects, the threshold for evoking the AMFR was within 0-25 dB of hearing threshold. It therefore appears that the AMFR may be a potentially useful tool to assess hearing in those unable to undergo conventional audiometric testing.