Neural structure of a sensory decoder for motor control

Abstract

We seek to understand the neural mechanisms that perform sensory decoding for motor behavior, advancing the field by designing decoders based on neural circuits. A simple experiment produced a surprising result that shapes our approach. Changing the size of a target for smooth pursuit eye movements changes the relationship between the variance and mean of the evoked behavior in a way that contradicts the regime of "signal-dependent noise" and defies traditional decoding approaches. A theoretical analysis leads us to conclude that sensory decoding circuits for pursuit include multiple parallel pathways and multiple sources of variation. Behavioral and neural responses with biomimetic statistics emerge from a biologically-motivated circuit model with noise in the pathway that is dedicated to flexibly adjusting the strength of visual-motor transmission. Flexible adjustment of transmission strength applies much more broadly to issues in sensory-motor control such as Bayesian integration and control strategies to optimize motor behavior.

Amplified cortical neural responses as animals learn to use novel activity patterns

“…One basic neural computation is representation, storing information about the sensory world in patterns of activity. Another is decision, or readout, in which representations are transformed or categorized by circuits into forms suitable for action (Egger and Lisberger, 2022; Wu et al, 2020). This readout process for object identification might, for example, lead to neurons in areas downstream from primary sensory cortex that encode category identity (Freedman et al, 2003), with neurons’ firing indicating whether a given sensory stimulus is food or stone; predator or friend.…”