Max Planck Institute for Dynamics and Self-Organization -- Department for Nonlinear Dynamics and Network Dynamics Group
Personal tools
Log in

BCCN AG-Seminar

Tuesday, 03.04.2007 16 c.t.

Timescale-dependent criticality of repeating spatiotemporal spike patterns

by Kai Gansel
from Max-Planck-Institut für Hirnforschung, Frankfurt/Main

Contact person: Raoul Martin Memmesheimer


Seminarraum Haus 2, 4. Stock (Bunsenstr.)


Timescale-dependent criticality of repeating spatiotemporal spike patterns The concept of self-organized criticality implies that systems of interacting nonlinear elements evolve over time into a critical state in which event sizes are scale-free and can be characterized by a power-law. Recently, this concept has been applied to neural networks, and circumscribed spatiotemporal patterns of local field potentials have been shown to meet the criteria of self-organized criticality and to fulfill many of the requirements expected of a substrate for information transmission and storage like high temporal precision and long-term stability. To investigate to which extent these findings may also be valid for the organization of spiking activity, we first detected repeating spatiotemporal firing sequences from rat cortical slices using sliding windows with lengths of 5 to 50 milliseconds. Additionally, longer sequences composed of temporally non-overlapping multineuronal spike patterns were assessed that lasted up to many seconds. Their significance was checked by common Monte Carlo methods. In this way, we were able to collect thousands of statistically significant spatiotemporal firing sequences from ten slices, occurring on different timescales that span several orders of magnitude. The distributions of their spatial and temporal dimensions were then searched for signs of criticality. Following the results, I will show that the neural system under study is far from being in a critical state concerning the spatial arrangement of repeating firing sequences, rather favoring well defined small cell assemblies. Likewise, sequences that had been restricted to 50 milliseconds at maximum exhibit approximately uniform distributions of durations, with the exception of a prominent peak around a few milliseconds. Longer sequences, however, display distributions of pattern durations that can be characterized by a power law, suggesting the existence of a critical branching process operating in the temporal dynamics of living neural networks on timescales well above 50 milliseconds. These findings will be discussed in the context of theories for neuronal coding.

back to overview