Reliable recall of spontaneous activity patterns in chaotic cortical networks

Irregular ongoing activity in cortical networks is often modelled as a result of recurrent connectivity. Yet it remains unclear how its presence corrupts the signal transmission evoked by the sensory drive. Here we show that reproducible responses in a generic recurrent cortical-like network can be obtained if the imposed external drive is consistent with patterns previously seen in the spontaneous, self-sustained activity. A subset of neurons in the network, constrained to replay a spiking pattern previously recorded during spontaneous activity, reliably drives the remaining, free-running neurons to reproduce the rest of the pattern. Comparison with surrogate Poisson patterns indicates that such transmission of input patterns is optimal for inputs with the statistical properties of the spontaneous activity. We propose that the similarity between spontaneous and evoked activity in sensory cortical areas could be the signature of learned efficiency of transmission across cortical networks.