Extended Data Fig. 4: Distinct neuronal sequences cover the choice to reward interval for left and right trials.

a, Schematic showing different trial types and phases of the task (choice and reward). b, For each session, the three cell types of interest, screen cells, reward-approach cells, and reward cells, are concatenated. Neurons are divided into two groups, population 1 and population 2, based on their responses to each trial type. Neurons with higher averaged peak activity during left trials are assigned to population 1, while those with higher averaged peak activity during right trials are assigned to population 2. Within each population, cells are sorted by the timing of their peak activity. The white dashed line indicates the choice time, and the black dashed line marks the onset of the reward. The time between choice and reward is scaled to one second. c, Task vs ITI neuronal responses for different cell types are plotted. The neuronal activity (averaged deconvolved calcium traces) for each of the cell types during the task vs ITI is plotted. For all cell types, there is a significant difference between task and ITI moments, demonstrating the task-related engagement of these cells. This is not observed for other cells (cells that are not identified as any of the three cell types). Top row: each point is one cell, and we have combined all sessions of one mouse (mouse #5) (n = 1,152 reward cells, 562 reward-approach cells, 686 screen cells, and 6,781 other cells). Bar graph and error bar in the insets show mean ± s.e.m. Bottom row: averaged across all cells for each mouse; each point represents the mean activity for one of the mice (n = 7 mice). The error bars on the points show s.e.m. Bar graph and error bar in the insets show mean ± s.e.m. In d–i, we show results for decoding task-related contextual features while conditioning on space. d, Task phase decoding (sample vs choice) during left screen contact (n = 181 sessions). Two-sided Wilcoxon signed-rank test P-value = 9.4979e-10. e, Task phase decoding (sample vs choice) during right-screen contact (n = 168 sessions). Two-sided Wilcoxon signed-rank test P-value = 4.3287e-13. f, Decoding correctness during left screen sample contact (n = 132 sessions). Two-sided Wilcoxon signed-rank test P-value = 0.0220. g, Decoding correctness during right-screen sample contact (n = 118 sessions). Two-sided Wilcoxon signed-rank test P-value = 0.0083. h, Decoding correctness during left screen choice contact (n = 131 sessions). Two-sided Wilcoxon signed-rank test P-value = 1.1285e-07. i, Decoding correctness during right-screen choice contact (n = 118 sessions). Two-sided Wilcoxon signed-rank test P-value = 3.2030e-06. Bar graphs and error bars in the bottom row of d–i show mean ± s.e.m. Each point represents one session. The illustrations in a,d–i were created using Affinity Designer.