Learning in neural circuits: insights from electric fish
Abstract: Understanding how learning is implemented in the brain is a challenging problem: in part because the synaptic plasticity underlying learning is embedded within intricate neuronal circuits the functions of which are often not well understood. I will discuss our efforts to combine physiological recordings, computational modeling, and most recently connectomics, to construct realistic models of an ecologically relevant form of learning in the electrosensory lobe (ELL) of weakly electric mormyrid fish. The ELL integrates somatotopic input from electroreceptors on the skin with a rich array of sensory and motor information conveyed by a mossy fiber-granule cell system similar to that found in the cerebellum. Prior work revealed how anti-Hebbian synaptic plasticity at granule cell synapses shapes motor corollary discharge signals into temporally-specific predictions that cancel out responses to self-generated input due to the fish’s own EOD. I will discuss recent work suggesting that the ELL performs more complex learning-related functions, including the prediction of sensory input related to the behavior of conspecifics. Finally, I will discuss our ongoing efforts to use connectomics to map the flow of electrosensory and granule cell inputs through the ELL. This analysis reveals highly-specific connectivity patterns that strongly support key predictions from prior electrophysiological and modeling studies as well as unexpected features that may provide a substrate for more complex forms of learning.