FIGURE 4. Analysis of reverse replay across all recording sessions.

a, For each session, a histogram of the rank-order correlation values of all events is shown in red, and a histogram of shuffled correlation values is shown in blue (see Methods). The two distributions were statistically different. P values for each session were as follows, where n is the total number of events. New: Rat 1, n = 1,425, P = 3.04 10^-41; Rat 2, n = 202, P = 4.00 10^-98; Rat 3, n = 91, P = 6.52 10^-10; Rat 4, n = 160, P = 4.88 10^-21. Familiar: Rat 1, n = 178, P = 6.69 10^-8; Rat 2, n = 33, P = 4.05 10^-6; Rat 3, n = 275, P = 0.0067; Rat 4, n = 88; P = 1.32 10^-8. The percentage of events with significant reverse correlations was as follows, by session: New, Rat 1, 13%; Rat 2, 72%; Rat 3, 31%; Rat 4, 29%; Familiar, Rat 1, 19%; Rat 2, 30%; Rat 3, 6%; Rat 4, 16%. b, Histogram of the number of significant reverse events per stopping period, for those stopping periods with at least one significant event. c, Cross-correlogram of significant reverse replay events with hippocampal sharp waves, for an example session in which there were 94 coincident events out of a total of 146 replay events.