We have studied a population of synfire chains above saturation. By considering a Hebbian-like learning strategy we show that to keep the propagation of activities controlled, the system reorganizes its structure leading to a new set of chains with shorter lengths distributed according to a non-trivial profile.