(A) Waveform properties of RSC border cells versus other cells recorded on the same tetrodes show no major difference between both populations. (B) Border cells were recorded both in granular and dysgranular layers of RSC across the recording depths. (C) An unbiased classification approach was applied based on linear-nonlinear models (Hardcastle et al., 2017). Three variables, Xi, were included: B, a one-dimensional vector of distance to the closest boundary, S, the animal’s running speed, and H, the animal’s allocentric head-direction. Models were built with increasing complexity using a forward-search approach (e.g., starting with one variable, then adding more if the model’s performance increases significantly, tested using ten-fold cross-validation). The model that best explained (e.g., had maximal log-likelihood) the cell’s firing rate, r, using optimal weights, wi, was then selected. Results show that boundary, speed, and head-direction encoding in RSC are independent features, as they can be expressed in isolation, combined with a substantial number of neurons that display conjunctive coding. (D-E) Tetrode cluster quality metrics that quantify the isolation distance (median = 15.39, IQR = 11.26–26.26) and L-ratio (median = 0.67, IQR = 0.17–1.51), based on Kilosort single vector decomposition (SVD) factor loadings (Schmitzer-Torbert et al., 2005). (F) There was no bias in the animal’s behavior around walls, as the running speed of the animals was uniformly distributed as a function of distance to the closest wall. (G) Both border cells and other cells recorded in RSC are modulated by running speed, with a ramping of firing rates as the running speed increases. However, border cells further showed lower firing rates in the low-speed range compared to other recorded cells. (H) Speed of the animal was uniformly distributed across the 2D space of the arena. *p<0.05, Wilcoxon signed-rank test (D), Wilcoxon ranksum test (E).