Taming a behavioral monster: Resonant song recognition and the evolution of acoustic communication in crickets

Abstract

AbstractRare behavioral phenotypes—behavioral monsters—can challenge hypotheses about the evolution of the neural networks that drive behavior. In crickets, the diversity of song recognition behaviors is thought to be based on the modification of a shared neural network. We here report on a cricket with a novel resonant song recognition pattern that challenges this hypothesis. Females of the speciesAnurogryllus muticusrespond to pulse patterns with the period of the male song, but also to song at twice the period. To identify the mechanisms underlying this multi-peaked recognition, we first explored minimal models of resonant behaviors. Though all of the three simple models tested (autocorrelation, rebound, resonate and fire) produced some kind of resonant behavior, only a single-neuron model with an oscillating membrane qualitatively matched the Anurogryllus behavior with regard to both period and duty cycle tuning. Surprisingly, the rebound model, a minimal model of the core mechanism for song recognition in crickets, fails to reproduce the preference for higher duty cycles observed in the behavior, questioning the universality of the core algorithm. However, the behavior is reproduced with a network model that contains all computations from the song recognition network of crickets, revealing the importance of an additional computation not part of the core mechanism: Following the core rebound mechanism in which post-inhibitory rebounds give rise to the resonant period tuning, feed-forward inhibition further shapes the tuning, resulting in the observed behavioral profile. Overall, this shows how unusual behavioral phenotypes can evolve by combining different nonlinear computations at the level of single cells and networks.