a, Spatial rate map of ten representative place cells are plotted. For each cell the left panel is the raw rate map and the right panel is the masked rate map (masking has been done by turning all bins with value less than 90% percentile to zero). Place cells are identified if their spatial information content exceeds 99% percentile of distribution of spatial information content for 1,000 shuffled cases. b, Distribution of spatial information content for place cells and non-place cells. Non-place cells: n = 1,519, 2,494, 1,979, 2,228, 4,706, 267, 175. Place cells: n = 1,003, 2,715, 2,661, 1,469, 4,270, 261, 201. Bar graphs and error bars show mean ± std. c, Percentage of identified place cells across mice. In average 47.5 ± 2.5 % of the cells were identified as place cells (n = 11, 11, 11, 7, 24, 23, 13 sessions for mice 1 to 7, respectively). Bar graphs and error bars in b,c show mean ± s.e.m. d, We calculated averaged place cell’s field sizes for a various amount of masking threshold from 75% to 95% percentile. After doing visual inspection we decided to use 90% percentile for masking spatial rate maps to calculate the size of the place field. This choice of threshold led to average place cell’s field size = 3.6 cm. n = 12580 cells. Error bar shows 95% confidence interval for each side of place-field size distribution. e, The correlation between the percentage of place cells with time (session) across mice is calculated. Two-sided Wilcoxon signed-rank test against zero: 0.6875. Bar graph and error bar show mean ± s.e.m. We do not observe a consistent trend across mice. f, Spatial rate maps and vectorized rate maps of eight representative cells are plotted. The vectorized rate maps clearly demonstrate the modulation of neuronal activity by the mouse’s behaviour, specifically its heading direction. g, Two dominant stereotyped behaviours in this task are the mouse running from the reward area to touch either the right or left screen and then running back to the reward port. A distinct population of neurons encodes different phases of each of these behaviours. Here, we visualize 10 place cells encoding different phases of the mouse running from the reward port to the right screen and then running back to the reward port. Tuning curves include spatial rate maps and heading direction tuning curves. h, The sequences of cells for flattened stereotyped behaviours of running to the left and right are plotted. Neuronal responses are obtained by averaging across trials within a representative session. Distinct subpopulations of neurons support each of these two behaviours, remaining silent during the other movement. i, GPFA dimensionality reduction70 on the entire population for these two patterns of behaviour reveals perpendicular neuronal trajectories in the neuronal latent space. j, Sample spatial rate maps for 5 representative cells are plotted. k, Scatter plot of peak rate maps. Each point represents the position of the peak activity of spatial rate map for one cell. l, Density plot of the scatter plots shown in k. m, Reward over-representation score for all the sessions of mouse #5 are calculated. The score is measured based on the overall representation shown in l. It is defined as the average of 10% spatial bins closest to reward normalized by the average of all bins. n, Reward over-representation across all mice is plotted. Each point shows one session. Consistent with previous studies18,71, our analysis of spatial coding reveals an over-representation of the reward area once the mouse has learned the reward location (n = 11, 11, 11, 7, 24, 23, 13 sessions for mice 1 to 7, respectively). Bar graphs and error bars in b,c show mean ± s.e.m.