Fig. 2: Population activity in RSC exhibits different levels of stability in correct and incorrect trials.

a Average normalized activity in the recorded population of neurons separated by their preferred context. For clarity, only comparisons between days 1 and 2, and days 1 and 4 are shown. Activity is cross-validated, sorted by latency using odd trials in day 1, and plotted using even trials for days 2 or 4. b Pearson’s correlation coefficient matrices showing the similarity of population activity along trial duration (averaged across 100 iterations of cross-validated population vectors constructed from the preferred responses in each context, as shown in Fig. 2a). Briefly, population vectors in one half of resampled trials in day 1 are compared to population vectors in the other half of resampled non-overlapping trials in day 1. For the other days, all resampled trials in day 1 are compared to all resampled trials in days 2 to 5. In addition, to compare vectors of the same length across days, the population is resampled to 1609 cells within each iteration (the lowest number of cells that meet criteria in all 5 days). c Population vector correlation coefficients along the diagonals in Fig. 2b (n = 100 iterations). Note the higher stability and similarity of population activity in correct trials. Top insets indicate time bins (in black) where the mean correlation in correct trials >95th CI of the correlation in incorrect trials throughout the trial. Solid line, mean; shaded area, 95% CIs. For all panels: s trial start, d decision point, e trial end.