Distributed Time-Domain Representations in the Birdsong System

Abstract

Vocal communication systems are common throughout the animal kingdom, and learned vocalizations such as human speech or bird song are of particular interest. How are vocalizations represented in the brain? Neural codes represent solutions to information-processing problems posed by an animal's behavioral requirements. In vocal communication, sounds are used as symbols, embodying certain parameters or features that have special significance for a given species. For example, the song of male crickets will attract females, but only if it embodies certain species-specific characteristics. Physical attributes that affect sound transmission through the acoustic habitat may affect the design of vocalizations, as in the case of long-distance communication, but the use of symbols suggests a certain arbitrariness as to the parameters of the symbols. Additionally, where vocal communication is invoked as part of reproductive behavior, it evolves under sexual selection, which can favor solutions that are not optimal in the engineering sense. The idiosyncratic, species-specific features of bird songs—essential diagnostic tools for bird watchers—provide examples of this type of system.