Neural Code & Bursts - NEUROSCIENCE

Phototransduction

Synaptic Imaging

Working Memory

Hippocampus

Neural Code

LTP & LTD

Click to Enlarge
It is quite common in neurophysiological studies of brain cells to assay cellular activation by monitoring how stimuli affect average spike frequency (firing rate). However, it has long been suspected that the "neural code" may not be carried exclusively by rate. We have explored two other factors that may be important determinants of the neural code. One factor is bursts which are typically groups of 3-5 action potentials that occur at 5msec intervals (200Hz). Special synaptic mechanisms, such as facilitation, could produce "burst detectors" that ignore isolated spikes.

To better understand the potentialities of burst coding, we have investigated the biophysical basis of burst production in simulated pyramidal cells. Our current work indicates that the biophysics of burst production has the consequence that bursts signal the slope of synaptic input. This is quite different from simple cell firing, which signals how far the input is above threshold, irrespective of slope. We have been able to find several examples in the brain where this slope detection mechanism appears to be utilized.

A second coding issue has to do with theta oscillations and whether the phase at which a cell fires during a theta cycle is important. Our modeling work on working memory is in fact built on the idea of theta/gamma coding in which different memory items become active at different phases of theta (i.e., in different gamma subcycles of the theta cycle). One opportunity for directly testing the idea of theta phase coding is in the hippocampus, where previous work from O'Keefe's lab indicated that place cells firing at different phases, depending on where the rat is in the place field. We have analyzed cell firing from large populations of simultaneously recorded place cells and attempted to determine the relative importance of rate and phase information. This could be accessed by reconstructing the animal's position from the ensemble of measured spikes. By comparing the reconstructed position to the actual position, we found that the reconstruction error was greatly reduced by taking spike phase into consideration. This work provides strong support for the utility of phase coded information.