Extended Data Fig. 8. Corner cells primarily correspond to an allocentric reference frame.
(a) Schematic for calculating egocentric corner bearing (red) using head direction (green) and allocentric corner bearing (blue) (Methods). (b) Behavioral data of allocentric corner bearing in square and convex-1 environments from one representative mouse. Each position is color-coded for the allocentric bearing of the nearest corner relative to the animal. Note the discrete color shifts represent changes in the closest corner to the animal (e.g. the northwest versus southwest corner). (c) A corner cell example with spikes color-coded according to the egocentric corner bearing in square and convex-1 environments. 0 degrees indicates the animal is directly facing the nearest corner. (d) Schematic of the linear-non-linear Poisson (LN) model framework with behavioral variables including allocentric position (P), allocentric head direction (H), linear speed (S) and egocentric corner bearing (E) (Model 1, see Methods). (e) True tuning curves (top) and model-derived response profiles (bottom) from an example corner cell. (f) An example of evaluating the model performance and selecting the best model using a forward search method. This example is from the corner cell in (e) and the best fit model (red dot) is the position (P) only model. (g) Number of corner cells (concave or convex) that were classified in each cell type category. This plot combined all the corner cells from all mice identified from the large square or convex-1 in Fig. 4 (a total of 77 corner cells (concave) and 44 corner cells (convex) from 10 mice). Some corner cells could not be classified potentially due to low spike rates (NaN). (h) Same as (g), but plotted using the proportion of total corner cells in each animal. For the box plots, the center indicates median, and the box indicates 25th and 75th percentiles. The whiskers extend to the most extreme data points without outliers (+).