Design and operation of the automated variant of the Barnes maze a) Scheme (left) and picture (right) showing a side view of the apparatus with its different components b) Scheme showing the operation of the apparatus. A new start position is set by rotating the arena. The goal box is simultaneously moved to align with the goal location.

Statistical characterization of vestibule sequences across days a) Top view of the maze (left) and two example trajectories on top of an overlay of multiple trajectories (right) showing vestibules and trajectory segmentation. Running segments span between two vestibule visits. Segment size is defined as the number of door-intervals a segment covers in the clockwise (positive sign) or counterclockwise (negative sign) direction. Serial bouts are bouts of consecutive 1-door-long segments. b) The average path length as a function of segment size for 3 time periods (lines) and across days (color coded). c-f) Same display as in (b) for the distribution of trial lengths (the number of segments per trial) (c), the distribution of visits across vestibules (d), the distribution of segment sizes (e), and the distribution of serial bout lengths (f).

Statistical characterization of vestibule sequences across trials a) The average path length as a function of segment size for 2 time periods (lines) and across trials (color coded). The data from days 6 to15 were used for this analysis. c-f) Same display as in (a) for the distribution of trial lengths (the number of segments per trial) (b), the distribution of visits across vestibules (c), the distribution of segment sizes (d), and the distribution of serial bout lengths (e).

Stochastic processes for random, spatial and serial strategies a) Schemes describing the stochastic processes for random, spatial and serial strategies. Vi+1: the next vestibule visited; Vi: the current vestibule; S: the next segment (a number of door-interval and a sign indicating clockwise/counterclockwise direction); and green lines: the probability distributions for random draws. Random strategy: Vi+1 is randomly drawn from the uniform distribution. Spatial strategy: Vi+1 is randomly drawn from a symmetric exponential distribution, and redrawn if Vi+1=Vi. Serial strategy: the sign of S is drawn with a 0.8/0.2 probability bias for positive/negative values. Then S is randomly drawn from a normal distribution with distinct center and width values depending on the sign of S, and is redrawn if S=0. Then Vi is incremented by S to obtain Vi+1. b) The same analyses as in Figure 2c-f, carried on vestibule sequences outputted by each stochastic process over 10 trials and 20 mice. For each trial, the stochastic process was recursively run until Vi=0. Note that none of the individual processes is reproducing the distributions of Figure 2.

Fit of vestibule visit patterns by a model combining random, spatial and serial stochastic processes a) Fits of experimental distributions for 5 days examples, using a model where strategy draws occur each n segments. Color coded, mean square error (m.s.e) of the fits for all combinations of random, serial and spatial strategies (note that P_spatial = 1 - P_random - P_serial). Line plots, overlays of experimental (blue) and model (red) distributions, for the best fits. b) Minimum mean square errors for the models where strategy draws occur each trial (black), each segment (blue), and each n segments (red), across days. c) The optimal values of n for the model in which strategy draws occur each n segments. d) Proportions of each strategy across days, obtained from the best fits.

Detailed view of the automated variant of the Barnes maze

(a-c) Pics of the entire apparatus (a) and of the side (b) and top views (c) of the apparatus.

(d) Zoom on the side doors.

(e-f) Zoom on the inside of one of the home boxes.

(g-h) Side view of one of the home boxes

Statistical characterization of vestibule sequences across days, for male and female separately

a) The average path length as a function of segment size across the 15 training days and 4 reversal days (color coded) and averaged across day 6 to 15 (lines) for male and female separately.

b-e) Same display as in (a) for the distribution of trial lengths (the number of segments per trial) (b), the distribution of visits across vestibules (c), the distribution of segment sizes (d), and the distribution of serial bout lengths (e).

Path length and number of segments per trial across days

a) Average path length per trial across days, for male (left) and female (right).

b) Average number of segments per trial across days, for male (left) and female (right). (mean±standard deviation, two-tailed paired t-test)

Quadrant and vestibule preference during the probe test

a) Fraction of time spent in each quadrant of the arena during the probe test, for male (blue), female (red), and all mice (black).

b) Fraction of time spent in each vestibule during the probe tests, for male (blue), female (red), and all mice (black). (mean±standard deviation, two-tailed paired t-test)

Vestibule orientation determines the direction of serial behavior

a) Top view of the arena (upper) and distribution of segment sizes (lower) for a leftward orientation of vestibules (red arrow).

b) Same as (a) for a rightward orientation of vestibules. Notice the reversal of the two peaks of the distributions, indicating that mice tend to visit consecutive vestibules via the clockwise (anti-clockwise) direction for the leftward (rightward) vestibule orientation.

Statistical characterization of vestibule sequences across trials, for male and female separately

a) The average path length as a function of segment size across trials (color coded) and averaged across trials (lines) for male and female separately. The data from days 6 to15 were used for this analysis.

b-e) Same display as in (a) for the distribution of trial lengths (the number of segments per trial) (b), the distribution of visits across vestibules (c), the distribution of segment sizes (d), and the distribution of serial bout lengths (e).

Models that combine random, spatial and serial stochastic processes Schemes of the models combining the random, spatial and serial stochastic processes. First, a strategy is drawn according to the set of probabilities P_random, P_serial and P_spatial. Second, the stochastic process associated with the selected strategy is used to draw the next vestibule. In model 1, the strategy is drawn at the beginning of the trial and maintained throughout the trial. In model 2, the strategy is redrawn for each segment. In model 3, the strategy is redrawn every n segments.

Fit of vestibule visit patterns and methodology comparison

a) Fits of experimental distributions across the 15 training days and 4 reversal days, using the model 3 where strategy draws occur every n segments. Color coded, mean square error (m.s.e) between model and experimental distributions for all combinations of random, serial and spatial strategies (note that P_spatial = 1 - P_random - P_serial). Line plots, overlays of experimental (blue) and model (red) distributions, for the best fits.

d) Proportions of each strategy across days, obtained from the best fits.

c) Criteria previously used to classify strategies in the Barnes maze. Trials with less than 3 vestibule visits are assigned a spatial strategy. Trials for which the goal is reached via a serial bout at least 3-door-long are assigned a serial strategy. The rest of the trials are assigned a random strategy.

d) Proportions of each strategy across days, calculated with the criteria described in (c).