Illumination mediates a switch in both active sensing and control in weakly electric fish

To execute sensory-guided behavior, the nervous system must manage uncertainty within multiple streams of information. There are two highly nonlinear mechanisms for achieving this: 1) sensory reweighting, an internal neural computation which places more emphasis on sensory information that exhibits the least uncertainty, e.g. in a Bayesian framework, and 2) active sensing, an overt behavior that seeks to improve the quality of sensory information before it enters the nervous system. Here we show that animals solve both of these nonlinear problems concurrently. We studied how the weakly electric glass knifefish Eigenmannia virescens alters its movement dynamics under parametric manipulations of illumination. We hypothesized a concomitant switch in both overt active sensing and internal multisensory reweighting. To test this, we varied illumination levels from 0.1 to 210 lx as fish tracked a moving refuge. We discovered that in a neighborhood of a critical threshold (on the order of 1 to 10 lx), small increases in illumination led to dramatic changes in both active sensing and multisensory control, specifically in 1) steep reductions in fish head and tail movements and 2) decreased refuge tracking phase lag. Outside of this threshold, large changes in illumination only caused small changes in active sensing and control. A control-theoretic model that dynamically modulates the weights of vision and electrosense due to illumination changes corroborates our findings. These findings underscore the complex, multipartite, nonlinear nature of locomotor control and the remarkable ability of the nervous system to execute multiple parallel strategies for managing sensory uncertainty.