There are two general neurophysiological models of sound lateralization mechanisms which may be active in man. Both of the models are derived from studies in animals (one in barn owls, and one in mammals), and both have displayed some weakness in generalizability. One model advocates a population of neurons narrowly tuned to different interaural disparity values across the behaviorally relevant range, so that the cue value, and therefore the source azimuth, is represented by which neurons of the array are activated by the stimulus. The second model posits the existence of only two neural channels, each broadly tuned to interaural cue values favoring one acoustic hemifield, so that, especially for sources near the midline, cue value and therefore source azimuth is encoded by the relative activation of the two neural populations. The present article reviews three recent psychophysical studies, each using selective adaptation paradigms to probe sound lateralization mechanisms based on interaural disparities in normal human listeners. These experiments provided evidence on the frequency-specificity of interaural disparity coding and revealed its sensitivity to recent stimulus history. The data from those studies, however, also help distinguish the two lateralization models, and favor a perceptual architecture for sound lateralization in man based on the activity of two, hemifield-tuned azimuthal channels.