Propagation of quasi-stable activation in a chain of recurrent neural networks

Abstract We consider a chain of feed-forward connected recurrent networks of leaky-integrate-and-fire neurons with membrane potential bistability. It is shown that lump neural activation, lasting for several seconds at each recurrent network, is gradually relayed along the chain. The time scale of the lump activation and its propagation along the chain agrees with that characterizing mental process. These results suggest that our model will provide possible neural mechanisms underlying mental representation of episodes.