Friday, July 30, 2004

12 noon

Redwood Neuroscience Institute

 

Title:   “Spike timing and visual coding in the LGN: How the cortex gets news of the world.”

 

Pam Reinagel

University of California

San Diego

 

Abstract:

We are interested the neural code used to transmit sensory information to cortex. Cortex may be responsible for the sophisticated computations that give rise to conscious vision, object recognition, and memory. But (as the data processing inequality guarantees us) all the visual information the cortex will ever have is embedded in the spike trains it receives from the thalamus. How is this primary information encoded?

 

We have found that the LGN encodes visual information in precise spike timing. There are several fundamental questions about any such code even at the level of single cells: how reproducible is the occurrence of each spike? how precise is the time of each spike? is each spike independent, or are patterns of spikes important? how universal or unique is the temporal code of different neurons? I will present the case that LGN spikes are reproducible from trial to trial, precise to order 1ms, patterned to at least order 10ms, and largely universal within a cell class. On the basis of this framework, I will outline some of the open questions we are now exploring. For example we are interested in how these coding properties depend on visual stimulus statistics; the significance of an intrinsic conductance that causes bursting; how coding properties change in the course of sensory adaptation; the cellular mechanisms for achieving temporal precision; the diversity and discreteness of neural types; and how to understand coding at the population level in the face of the curse of dimensionality.