In addition, while single-cell activation probabilities and the

number of cell pairs active were numerically higher before correct than incorrect trials, neither of these factors could fully account for the measured pairwise differences (Figures S1D and S1E). These findings suggest that specific sets of cell pairs were strongly activated before correct trials, a possibility we confirm below. We have previously shown that coactivation probability during SWRs is high in novel environments and then decreases with experience (Cheng and Frank, 2008). Here we found that there was greater relative coactivation probability preceding correct, as compared to incorrect, trials for 65%–85% and >85% correct performance categories. We therefore sought to understand how trans-isomer chemical structure differences in coactivation probability between correct and incorrect trials interacted with the overall decrease in coactivation probability with experience. We combined data across tracks and found that coactivation probability preceding correct trials remained high from the first exposure through the first session with >85% correct performance (Figure 2D; Capmatinib order p’s > 0.1 for comparisons among correct trials for <65% correct, 65%–85% correct, and >85% correct performance categories). In contrast, coactivation

probability dropped significantly for incorrect trials during learning (65%–85% correct and >85% correct, p’s < 0.001 versus <65% correct). Finally, once animals achieved consistent >85%

correct performance, coactivation probabilities dropped for correct trials (p’s < 0.001 versus <65% correct, 65%–85% correct, and >85% correct) to a level similar to that seen on incorrect trials. These findings suggest that errors made during learning reflect lower levels of place cell pair coactivation during SWRs. The lower mafosfamide levels of coactivation probability on incorrect trials also account in large part for the differences in Z scores before correct and incorrect trials. We computed the mean difference in coactivation probability for each pair, defined as the mean coactivation probability on correct trials minus the mean coactivation probability on incorrect trials. Not surprisingly, this coactivation probability difference was strongly correlated with the Z score measure (r = 0.85, p < 10−4). This indicates that large differences in coactivation probability for individual pairs is a strong driver of the Z score effects, with the remaining variability in the Z scores arising from the influence of the different numbers of SWRs before correct and incorrect trials. We then asked whether incorrect or correct trial coactivation probability alone was a better predictor of Z score. We found that coactivation probabilities on incorrect trials for individual cell pairs were significantly negatively correlated with the Z score measure for those pairs (r = −0.59, p < 10−4). Thus, low coactivation probability predicted high Z score differences.