Extrinsic and intrinsic dynamics in visuomotor tracking

Abstract

Humans typically produce 2–3 submovements per second when tracking slow targets. This intermittency is altered by the addition of delays in sensory feedback suggesting that it is governed by extrinsic properties of the control loop. However, the motor cortex also exhibits an intrinsic rhythmicity at 2–3 Hz, which might influence the temporal structure of movements. This thesis examines how the interplay between extrinsic and intrinsic dynamics shapes the kinematics of tracking behaviour. I found that the dependence of submovement frequencies on extrinsic delays could be reproduced by a simple feedback controller model. This model predicted that submovements reflect frequencies at which visuomotor noise is exacerbated, and this was confirmed by perturbation experiments. However, these experiments also revealed a 2-3 Hz band-pass filtering of feedback responses irrespective of extrinsic delay. Further experimental evidence suggested this filter did not reflect properties of either visuomotor noise, the feedforward pathway, or visual processing. However, the filter exhibited features consistent with a state estimator required for optimal feedback control (OFC) in the presence of visual and motor noise. Finally, I sought evidence that this filter was implemented by motor cortical circuits. Multichannel local field potentials (LFPs) in the motor cortex of macaque monkeys were strongly correlated with submovements, at frequencies which depended on extrinsic delay. However, the dynamics of LFP cycles during submovements were independent of delay, and matched instead the properties of the state estimator in the OFC model. In summary, by combining human behavioural studies, computational modelling and monkey electrophysiology, I show how movement intermittency can be explained by the interplay of both extrinsic and intrinsic dynamics within an OFC framework. Moreover, I suggest that motor cortical rhythmicity reflects recurrent circuitry that combines sensory feedback with an internal dynamical model to form optimal estimates of required motor corrections.