Seeking neural signatures

Stephen Cowen1, Richard S. Zemel2, Jason Gerrard1, Hemant Kudrimoti1 and Bruce McNaughton1

1University of Arizona
2University of Toronto

Hebb's cell assembly hypothesis proposes that the associative nature of neural connectivity allows ensembles of neurons to form functional groups. Since then, numerous experimental and theoretical investigations have suggested a neural code consistent with this idea, codes far more rich in structure than a coarse rate code. All of these studies have at their core the common idea of what we term a neural signature, a code for a mental event involving a detailed temporal firing pattern in a group of cells.

Many of the techniques used to investigate these coding strategies require the neural activity to be tightly linked to particular stimulus attributes. However, in the case of cognitive processes such as decision making or planning, the temporal correlation between neural activity and stimulus attributes may vary considerably from trial to trial. We have developed a method that can potentially identify salient features in neural population recordings regardless of when those features occur. The model can also suggest when during a trial a decision was made. We apply this model to population recordings of rats doing a simple maze exploration task, and compare the results to standard pattern recognition approaches.