Can a fish learn to ride a bicycle? Sensorimotor adaptation to destabilizing dynamics in the weakly electric fish Eigenmannia virescens

1Humans and other animals can readily learn to compensate for destabilizing dynamics, such as balancing an object or riding a bicycle. How does the nervous system learn to compensate for such destabilizing dynamics, and what are the benefits of the newly learned control policies? To investigate these questions, we examined how the weakly electric glass knifefish, Eigenmannia virescens, retunes its control system in the face of novel, destabilizing dynamics. Using a real-time feedback system, we measured swimming movements as seven individual fish tracked a moving refuge, and we fed the swimming movements back through novel dynamics to alter the refuge motion, creating an artificially destabilizing reafferent loop. We discovered that fish learned to retune their sensorimotor controllers as the artificially destabilizing feedback was gradually introduced. Furthermore, when the artificial feedback was extinguished, fish exhibited a clear aftereffect, retaining their learned sensorimotor controllers for several minutes before washing out. This retuning of the control system under destabilizing dynamics: (i) improved tracking performance compared to the predicted performance had fish not re-tuned their baseline controller, (ii) reduced sensitivity of the sensorimotor system to low-frequency disturbances, such as would arise from turbulence or motor noise, and (iii) improved phase margin, a measure of stability robustness, despite the artificial feedback driving the putative baseline control system towards instability. Our study sheds light on how the nervous system adapts to changing closed-loop dynamics, and how those changes impact performance and stability; the presence of aftereffects suggest a plasticity-based mechanism reminiscent of cerebellar learning.