Extended Data Fig. 5: Homing runs, turn angles, and heading directions.
From: Mice learn multi-step routes by memorizing subgoal locations
a, Left: histogram of homing runs’ initial condition. This shows, for an average escape in the CORE-ZB, how many prior homing runs fell into different proximity bins. Each bin reflects proximity in both the position (x-axis) and body orientation (y-axis) of the homing’s starting point. Right: example of homing runs extracted from exploration during the CORE-ZB’s exploration period, and example of a subsequent escape in that experiment. b, Experiment with narrow corridors that constrain movements during exploration and escape. Histogram and homing runs are computed in the same manner as in panel a. c, Correlation between the predicted escape target (using the procedure illustrated in Fig. 3b) and the actual escape target. Here, escape targets are predicted using the homing run with the most similar turn angle to the escape turn angle. Data are from homings and escapes in the platform with narrow corridor. The correlation coefficient r = 0.98; P = 2×10-22. This prediction thus yields an R2 value of 0.97, which here corresponds to a mean absolute error of 0.025 in escape-target units. Note that we are using R2 values as a measure of explained variance and not as a measure of goodness-of-fit of the linear relationship. d, Same analysis as in panel c, but with escapes and homings from the CORE-ZB. The correlation coefficient r = -0.03; P = 0.9. This prediction thus yields an R2 value of 0.0009, which corresponds to a mean absolute error of 0.38 in escape-target units. For comparison, predicting that every escape will be equal to the CORE-ZB’s mean escape target generates a mean absolute error of 0.31. e, Correlation between the allocentric heading direction required to target the obstacle edge ___location (x-axis) and the allocentric heading direction during escape (y-axis). The y-axis heading direction is measured when the mouse is 15 cm away from the escape initiation point (that is, after the initial turn movement is complete). The vector from the center of the platform to the shelter (pointing south) is set as 0˚, and a vector pointing west or east is ±90˚. The absolute value of the heading direction is taken so that escapes toward the left and right edges can be considered together. Homing-vector escapes are not included. Data are from the two COREs of Fig. 2a,b. Lines show the linear regression fit, and the shaded area shows the prediction interval within 1 standard deviation.
