Information Processing in the Higher Parts of the Auditory Pathway

Abstract

The acoustic communication of grasshoppers, crickets, and katydids provides prime examples of general principles of how nervous systems represent sensory information. The processing of auditory signals in insects is innate; thus the involved neuronal modules are “hardwired.” In addition, the relevant stimulus space is restricted, facilitating the investigation. A major problem of sensory processing is the trial-to-trial variability of spike trains caused by the stochastic opening and closing of ion channels. In animals that can spend only a few neurons for a given task, this unreliability of spike trains is a relevant constraint for neuronal encoding. Signal recognition in insects depends primarily on features of the sound envelope, the pattern of amplitude modulations. The receptor neurons respond with high temporal precision and reflect the stimulus's envelope in their spike patterns. This kind of “temporal code” is later transformed to a “labeled line code” representation in which single neurons encode specific sound features. Similarly to the larger nervous systems of vertebrates, at this stage, spike rates are reduced and the presence of particular sound features can be read out from a population code. Remarkably, in insects, the sparsening and change of coding schemes occur already within a few synapses after the receptors. Modeling studies suggest how feature detectors equipped with linear filters may explain behavioral scores found with specific stimulus variations. Remarkably, in grasshoppers and crickets, the filters found by this approach resembled Gabor filters, which allow an easy transition between behavioral preference functions found in different species.