Discrimination of noise-contaminated communication signals by grasshoppers

Abstract

When using acoustic signals in order to find—and possibly select between—mates, grasshoppers face two major problems: the arriving signals are usually degraded by various kinds of noise, while as much information as possible should be extracted quickly. Major cues for signal recognition are the temporal patterns of amplitude modulations of grasshopper songs. Within the nervous system, signal classification is rooted in the spike trains provided by the sensory neurons and will be constrained by the variability of spiking responses. Several intrinsic sources further contribute to spike-train variability, which can be considerable even in the case where undistorted acoustic signals arrive at the receiver. Using van Rossum’s [Van Rossum, M. Neural. Comp. 13, 751–763 (2001)] method to compute spike-train distances, we quantified the variability of neural responses and showed that the temporal information necessary to discriminate between songs of different conspecifics is preserved surprisingly well, even in the responses of single neurons. By comparing behavioral and neurophysiological data, we now aim at determining how well this information can be used by the animals, and how closely or remotely their nervous systems approach the performance of an ideal observer.