Figure S7.

Replay Events in Long Epoch, Inter-Replay-Interval, and Outcome Reactivation, Related to Figure 7

a, b, Time-frequency maps shown for an extended epoch after replay events in study 1(a) and 2 (b). 0ms is the onset of replay events. Notably these are only replay events which are not preceded by other replay events in the previous 100ms. Note that the increase in ripple-band and low frequency power extends for several hundred milliseconds. This can be explained by the fact that replay events occur in clusters. Histograms show the inter-replay intervals across all replay events in all subjects in study 1(a) and study 2(b). The modal replay onset time is immediately following the previous replay event. The heavy-tail of the distribution indicates that there are also periods with no replay events. c, To find the peak time point to train outcome decoders in the value learning phase from Study 1, we trained classifiers on every time point relative to the onset of outcome, and tested at every time bin relative to the same onsets. Each cell of this grid shows cross-validated (leave-one-out) prediction accuracy. We found the peak decoding accuracy was around 200 ms after the stimuli onset. d, The reward outcome reactivated at the same time as the onset of replay of the rewarded sequence during rest, while no such relationship was observed for neutral outcome and replay of neutral sequence.