The ability of the auditory system to cope with temporal subsampling depends on the hierarchical level of processing

Evidence for rhythmic or 'discrete' sensory processing is abundant for the visual system, but sparse and inconsistent for the auditory system. Fundamental differences in the nature of visual and auditory inputs might account for this discrepancy: whereas the visual system mainly relies on spatial information, time might be the most important factor for the auditory system. In contrast to vision, temporal subsampling (i.e. taking 'snapshots') of the auditory input stream might thus prove detrimental for the brain as essential information would be lost. Rather than embracing the view of a continuous auditory processing, we recently proposed that discrete 'perceptual cycles' might exist in the auditory system, but on a hierarchically higher level of processing, involving temporally more stable features. This proposal leads to the prediction that the auditory system would be more robust to temporal subsampling when applied on a 'high-level' decomposition of auditory signals. To test this prediction, we constructed speech stimuli that were subsampled at different frequencies, either at the input level (following a wavelet transform) or at the level of auditory features (on the basis of LPC vocoding), and presented them to human listeners. Auditory recognition was significantly more robust to subsampling in the latter case, that is on a relatively high level of auditory processing. Although our results do not directly demonstrate perceptual cycles in the auditory domain, they (a) show that their existence is possible without disrupting temporal information to a critical extent and (b) confirm our proposal that, if they do exist, they should operate on a higher level of auditory processing.