Fig 3. Positional error as a function of time from the teleportation light switch.

A. Recorded data. Positional error computed as a function of time from the teleportation (in units of theta cycle). In each theta bin t, the map decoder is used to select which map m_t (set of place fields) to use to infer the position from the neural activity (Fig 1A). When using the decoded map m_t, the positional error (blue line) is comparable to the one found in constant-environment conditions (blue dashed line). The orange line shows the positional error if the opposite map (alternative to m_t) is used for inferring position. The positional error is significantly reduced with respect to constant-environment conditions (orange dashed line) in the vicinity of the teleportation, and reaches comparable values after ∼ 80 theta bins. Shaded areas represent the standard error computed over 15 teleportation events. B. Recorded data. Same as in panel A but using a fixed map of reference for position inference, irrespectively from the decoded map. Red and green lines show results with, respectively, the ‘old’ (pre-teleportation) and the ‘new’ (post-teleportation) maps. C & D. Simulations of CANN model: same analysis as in panels A & B, computed over 15 simulated teleportation events, with the same trajectory of the rodent as in the experimental data. Model parameters: N = 400 neurons, γ_J = 0.0025, γ_V = γ_PI = 0.4, γ_W = 6.25, β = 15, see Methods and Figs. B&C in S1 Text for detailed discussion of the choice of parameters.