In the companion paper [V.d. Drift et al.; Audiology 27: 260-270, 1988], it was shown graphically that conductive and cochlear hearing loss can be distinguished on the basis of the combinations of the auditory brainstem response threshold with the horizontal shift of the latency-level curve of peak V, its derivative or the latency of peak V at threshold level, respectively. In addition to the patient data used in the companion paper, 22 patients with mixed hearing loss were enrolled in the present study. The statistical technique of discriminant analysis was applied to find the optimum linear combination of auditory brainstem response data for classification of a hearing loss. The brainstem classification 'cochlear hearing loss' agrees with the diagnosis on the basis of the pure-tone audiogram in 85% of the cases. In cases with the brainstem classification 'conductive hearing loss', 93% showed at least a conductive component in the pure-tone audiogram.