Principal Supervisor: Dr. David Souto, Department of Neuroscience, Psychology and Behaviour
Co-supervisor: Dr. Claire Hutchinson, Department of Neuroscience, Psychology and Behaviour & Dr. Emmauel Prempain, Department of Engineering
PhD project title: Computational principles of oculomotor plasticity
University of Registration: University of Leicester
The control principles and physiology governing smooth pursuit eye movements are relatively well known. Those movements are simple enough to be modelled and complex enough to address a number of questions about motor control and learning. Although we have relatively good models of how the brain controls smooth pursuit eye movements, we know little of its plasticity.
We will use a learning paradigms in which we will train participants to perform a novel eye movement task—i.e. draw a target pattern (a cursive “e”) with a cursor, which location changes according to an on-line estimate of eye position. By comparing response changes pre- and post adaptation to the predictions of computational models of pursuit control, we will be able to test specific hypothesis of how learning takes place.
First, interpreting visual feedback to control pursuit eye movements requires that the brain holds an accurate forward model of the effect of motor commands. This is important given processing delays, and hence the inability to control action based on outdated information. One mechanism for coping with those delays is by adjusting the forward model when feedback arrives earlier or later than predicted. By delaying visual feedback we can test whether the pursuit system is able to adapt to those changes. A misperceived feedback would predict a change in the temporal frequency of eye movements towards a moving target. We will also test whether motor adaptation is followed by sensory recalibration, as in adaptation to asynchronous percepts.
Second, it is typically assumed that oculomotor learning occurs via supervised learning, i.e. an error signal is used to guide the change in behaviour. However, reinforcement (reward) learning is another possibility. Recent simulations indicate that pursuit behaviour could have evolved by maximizing image sharpness for moving objects, providing an implicit reward. We will then test learning rates and learning generalization with and without explicit rewards, as a way to test whether unsupervised learning has an important role in pursuit plasticity.
In this project experimental research and modelling will go hand in hand, as we will use control systems models of pursuit eye movements to test hypothesis about the influence of motor learning (e.g. delay adaptation) on movement dynamics recorded with a state-of-the-art eyetracker. In particular we are interested in testing and improving on models of pursuit eye movements that apply to unpredictable target inputs. In those models the pursuit sensory-motor transformation is modelled by a series of filters, taking into account visual processing, feedback delays, motor integration stages and plant dynamics (e.g. viscoelastic forces) (Robinson et al. 1986; Orban de Xivry et al., 2013). For instance, adaptation to feedback delays would produce a specific shift in ringing frequency around the target velocity post-adaptation. The models will be adapted to incorporate the effect of reward on sensorimotor adaptation.
Perform computer simulations using Simulik (Matlab, Natick, US).
Dr. Emmanuel Prempain will assist in the modelling and experimental design, providing expertise in the mathematical foundations of control systems.
Dr. Claire Hutchinson will contribute expertise in visual psychophysics and perceptual learning.
In particular, the student will learn to:
- Program experiments requiring on-line data processing (gaze-contingent stimulation), and state-of-the-art eye-trackers - Analyse eye-movement data
- Build computational models of eye movements and refine them by comparing them to data
- Work within an interdisciplinary team
- Robinson, D. Ar, J. L. Gordon, and S. E. Gordon. "A model of the smooth pursuit eye movement system." Biological cybernetics 55.1 (1986): 43-57
- de Xivry, Jean-Jacques Orban, et al. "Kalman filtering naturally accounts for visually guided and predictive smooth pursuit dynamics." The Journal of Neuroscience 33.44 (2013): 17301-17313.
BBSRC Strategic Research Priority: Molecules, cells and systems
Techniques that will be undertaken during the project:
- Eye-tracking (Eyelink 1000)
- Statistical analysis (functions, R or Matlab)
- Psychophysical modelling (psychometric functions, Matlab)
- Computer simulations (Simulink, Matlab)
Contact: Dr David Souto, University of Leicester