SERIES EXPANSION ANALYSIS OF CORTICAL SPIKE TRAIN INFORMATION

 

 

   1Simon R. Schultz

2Stefano Panzeri

Howard Hughes Medical Institute & Center for Neural Science

1New York University, New York

2Department of Psychology

University of Newcastle Upon Tyne

 

Any direct estimate of the complete spike train information is limited by sampling considerations to relatively small word lengths, and therefore to the analysis of short time windows of data.  The series expansion approach takes advantage of this limit to create a more manageable estimation problem.  In this talk I will discuss the sampling issues with respect to a selection of primate V1 complex cells from which sufficient experimental trails were available.  Estimation of the noise entropy by both series and ‘brute force’ procedures will be compared.  We have also applied this analysis procedure to data recorded from the rat barrel cortex by M. Diamond and colleagues.  This analysis demonstrates that, in the whisker representation of rat cortex, precise timing of early spikes increases the information transmitted about stimulus location by an average of 44 percent beyond that transmitted only by the number of spikes.  This result stands in contrast to that observed in the coding of grating orientation by the V1 complex cells.