The potential usefulness of the auditory-evoked brain stem potentials in audiological assessment has been the subject of increased clinical enquiry. Several requirements need to be met by any response parameter if it is to be clinically useful: parameters should show a clear, and ideally simple, relationship with stimulus intensity; they should be stable within any subject under fixed stimulation conditions and intersubject variability within any clinical group should be small. Response latency has been shown to satisfy the first two of these conditions. However, intersubject variability has been less clearly defined. In a sample of normal controls we have found that the standard deviation of latency at any specific intensity was substantial, exceeding the difference found between mean responses at neighbouring intensities of auditory stimulation. In an investigation of the origins of this variability we have found a clear gender difference, maximal for later components in which the male responses were significantly later than the female responses. The clinical significance of this finding can be seen in the latency intensity input/output function. By treating each gender group separately, intersubject variance values were reduced by approximately 40% compared with the total group variance, so enhancing the usefulness of latency as a parameter in the objective assessment of auditory function.