(A) Characteristic examples of the network activity showing spiking events during stimulations of each individual ‘letter’ of a memory sequence in awake. (B) Distributions of the differences in response delay for all neurons from the trained region when the respective left vs right neighboring groups are stimulated, as shown in A. (C) Response delay-based directionality index before/after training of each sequence (S1/S1*) in gray, and after sleep (red). (D) 3-D surface plot showing, for each neuron from a trained region, the number of incoming synapses demonstrating reliable replays during sleep (z-axis), mean response delay during testing (as in panel A) after sleep (y-axis), and total synaptic input to a neuron after sleep (x-axis). Note that neurons receiving highest total synaptic inputs in specific network direction after sleep are also those who respond with shortest delay during testing recall in that direction after sleep and also those who receive the highest number of synapses demonstrating reliable replay during sleep. (E) Simplified cartoon of the network connectivity after different training phases followed by sleep. Arrows indicate connections between neurons (nodes) with blue arrows being connections strong for S1 and red for S1*. Blue and red nodes represent neurons that contribute (weakly - light colors; strongly - dark colors) to recall of S1 and S1*, respectively. Top, network configuration after S1 training – all nodes and connections are allocated to S1. Middle/Top, network configuration after initial S1* training – nodes/connections start to learn S1* and ‘unlearn’ S1. The information about the old memory S1 is still available. Middle/Bottom, network configuration after continuing S1* training – all nodes/connections are allocated to S1*. All information about S1 is lost. Bottom, network configuration when initial S1* training is followed by sleep – orthogonalization of memory traces, some nodes/connection are allocated to S1 and others to S1*.