Auditory evoked potentials generated in the cochlea and the brainstem can be recorded from the human scalp. These potentials generally have a latency of less than 15 ms and an amplitude of less than 2 microV. They are, nevertheless, relatively easy to record using averaging, provided that appropriate attention is paid to the frequency band-pass of the amplifier-filters and to the locations of the recording electrodes. The responses are precisely determined by a variety of stimulus factors such as intensity and frequency, and by many subject parameters such as age and sex. This paper considers the anatomical and physiological principles underlying these relationships.