Trial-by-trial variability of spike trains in vivo and in vitro
M.P. Nawrot, V. Rodriguez, D.Heck, A. Riehle (1), A. Aertsen, S. Rotter
Neurobiology, Biology III, Albert-Ludwigs-University, Freiburg, Germany
(1) CNRS-CRNC, Marseille, France
Cortical neurons observed
in vivo are known to produce highly irregular spike trains and a considerable amount of variability across trials. Both phenomena are related and have great impact on theories of neural coding that must be compatible with these conditions. Here, we compared
in vivo single-unit recordings from cells in M1 of awake behaving monkeys and
in vitro patch recordings from rat neocortical cells during injection of currents mimicking synaptic input under different controlled conditions. To estimate the variability across trials we employ the Fano factor, FF, i.e.~the ratio of count variance and mean spike count. The irregularity of spike trains is quantified by the coefficient of variation, CV, of inter-spike intervals. The statistical properties of both measures were numerically determined for a large class of renewal processes.
In vivo, the distribution of FF revealed systematic changes during task performance. Trial-by-trial variability was lowest during movement execution with a mean FF of 1, and highest during the initial waiting period before information about the target direction was provided. The results for a combined analysis of FF and CV did not correspond with the expectation for a time-modulated renewal process. In contrast, the
in vitro results were in very good agreement with this expectation. Our findings suggest that a considerable amount of the trial-by-trial variability
in vivo must be attributed to ongoing background activity. A simple model combining a deterministic time-varying rate with an additional non-deterministic level of background activity can explain most of our results.
Supported by DFG (SFB505), GIF, HFSP and DAAD
