Sensorimotor Computation Seminar Series - Talk by Ed Keller

SENSORIMOTOR COMPUTATION SEMINAR SERIES

http://sensorimotor.pwias.ubc.ca

Presents

Prof. Ed Keller
Professor Emeritus
University of California, Berkeley

A Perspective on the Oculomotor System: Neurophysiology and Computational Models

Monday 11 April, Refreshments at 12:30pm Talk 1:00pm - 2:30pm
Room X836, Institute for Computing, Information and Cognitive Systems/Computer Science
2366 Main Mall, UBC, Vancouver

ABSTRACT: The oculomotor system has served as a model system for the clarification of how the brain controls precision movements. It is a "one joint" system with rather simple muscles and mechanical properties. Deeper understanding of this system began with the work of David Robinson who showed, in contrast to long prevailing beliefs that the orbital mechanics could be modeled as an underdamped system, the system was heavily overdamped, i.e., it was dominated by viscosity. Based on this fact he predicted that the optimal neural signals from motor neurons to eye muscles would consist of a high frequency burst of activity that dropped to a much lower level of activity just as the eye reached a new eccentric position. The low level of activity then held the eye fixed at the eccentric position against the spring-like forces of the eye muscles. At about the same time a number of neuroscientists developed techniques to record individual neurons in the alert, behaving monkey. Robinson and others then recorded individual motor neurons in the monkey as these trained animals made saccadic eye movements to projected visual targets. Robinson's hypothesis was confirmed, and the motoneuron signals became known as pulse-step controls. Neuroanatomists showed which brainstem regions contained the second-order neurons that projected to motoneurons. Prominent among these areas was a region in the central reticular formation named the PPRF. This information made it possible to find and record from neurons that projected to motoneurons. Explaining how the signals in these upper motoneurons activated motoneurons, led to a Robinson model that has been the basis for further modeling efforts that still continue to the present time. Basically the neural signals at this level can be modeled with single, lumped model neurons and the coding is all temporal. The next advances came when neuro-scientists began to record from higher level neurons projecting to the PPRF (the superior colliculus and cerebral cortex). Here a major conceptual obstacle appears. Neurons at these higher levels are spatially, not temporally coded. This means that distributed models are required to make further progress on explaining the system organization. Efforts in this direction are being undertaken, but it is difficult to prove uniqueness and causality with these models.

BIOGRAPHY:

Edward Keller received his BS degree in engineering science from the US Naval Academy in Annapolis and his PhD in bioengineering from the Johns Hopkins University in 1971. He joined the faculty in Electrical Engineering and Computer Science at Berkeley in 1971. He became associated with the Smith-Kettlewell Eye Research Institute in 1979. In 1994 he became Professor Emeritus at Berkeley and then devoted full time to his research interests as senior scientist at Smith-Kettlewell. He served as Associate Director at this institute beginning in1998. He is a Fellow of the IEEE. His principal interests are modeling and neurophysiological studies of distributed processing in the oculomotor control system.

To Schedule a meeting with Prof. Keller, please contact Margaret Bloomquist Margaret.bloomquist@pwias.ubc.ca Office: 604.822.8218