Spike sequences and their consequences

Spatio-temporal patterns of spikes have an advantage of representing inform ation by their spike composition similar to words of languages. First we re view the models of neuronal coding, then we discuss technical aspects of de tecting spatio-temporal spike patterns. We argue by presenting data from ra t hippocampus that spike trains recorded simultaneously from multiple pyram idal cells are not independent. Their hidden dependency structure can be re vealed by spike 'sequences', defined as a set of neurons which fire in a sp ecific temporal order with certain delay between successive spikes. The onl y way to prove their existence in vivo is to show that they recur with high er than by-chance frequency. We observed that 'sequences' possess 'composit ional' features and that a given spike composition is time scale invariant. We illustrate that the same neuron can be a part of different 'sequences' and 'sequences' recur in a temporally compressed fashion during slow wave s leep. The statistical significance of 'sequences' is testable. Their biolog ical significance has been implicated by experiments where recurrence rate of the sequences during different behavioral sessions were compared. As con sistent with the 'replay hypothesis' of memory consolidation, new sequences generated during the wake state are persistent during the subsequent sleep . Thus, information acquired during the wake state and represented by spati o-temporal patterns of spikes may transfer to the neocortex: during sleep. Our results suggest that 'sequences' reflect the activation of specific but configurable circuitries during exploratory behavior, followed by spontane ous re-activation of the same circuitry during sleep. Whether the delay str ucture of spikes as a combination is an effective input to single neurons d ownstream or 'sequence' components are being processed in parallel pathways and evaluated independently is an open question. (C) 2000 Elsevier Science Ltd. Published by Editions scientifiques et medicales Elsevier SAS.