Behaviors replicated between cohorts.

a. Experimental timelines for cohort 1 (top) and cohort 2 (bottom). Phases include Linear Track (light green; LT), Novelty (dark green; NOV), Attack (red; ATTK), Extinction (EXT; blue), Re-exposure (orange; ReEX). Each block in the timeline represents one session (1 session per day, 1 hour in length). b. Sideview track schematic when robot is present. Rat runs between nest and far feeder to obtain food rewards, crossing in front of the robot. Attacks only occur during ATTK and ReEX phases, only on outbound journeys (black arrow). Inbound journey shown in grey. For analysis, different regions on the track are shown in different colors, Nest (purple), Track Start (blue), Attack (green), Robot (pink), Far Feeder (yellow) but are not physically present on the actual track. c. Example linearized trajectory on an attack day (example from R823-2023-08-06). Horizontal boxes of different colors represent regions. Landmark positions are shown using solid horizontal lines (doorway in grey, attack threshold in orange, robot in black). Red dashed vertical lines show when attacks occurred. Highlighted sections show behaviors of interest (retreats, MTCs, MTAs). d. Number of laps, MTCs, and MTAs across sessions and phases. Top row: cohort 1 (blue); bottom row: cohort 2 (red). Points represent individual animals, and square markers indicate the mean ± SEM across animals, with marker color denoting phase. For clarity, extreme outliers are not displayed, but were included in calculation of the error bars. Specifically, 2 outliers were omitted from the laps panel in cohort 1; 2 outliers from the MTC panel in cohort 1 and 3 in cohort 2; and 2 outliers from the MTA panel in cohort 1 and 3 in cohort 2.

Behavior probability.

a. Probability that an outbound pause resulted in an abortive decision (MTA) across sessions in the ATTK and ReEX phases, calculated as MTA / (MTA + MTC). One outlier from cohort 2 is not shown for visual clarity, but was included in the error bar calculation. b. Proportion of attacks that resulted in retreat behavior (retreats / attacks) across sessions of the ATTK and ReEX phase (when robot attacks occurred).

Speed dynamics.

a. Relative inbound (IB) speed across sessions and phases, relative to linear track (LT) baseline. Top row: cohort 1; bottom row: cohort 2. Points represent individual animals (color-coded by cohort), with square markers indicating the mean ± SEM across animals (color-coded by phase). b. Same as above but for outbound (OB) speed.

Speed differences between behavior.

a. Mean inbound speed associated with retreats, MTAs, and matched inbound control laps during ATTK. Top row: cohort 1; bottom row: cohort 2.

Comparison of inbound journey following retreats and MTAs.

a. Aerial-view schematic of inbound journey following a retreat (left) and mid-track abort (MTA; right). Retreats are triggered by robot attack (red arrow), whereas MTAs follow anticipatory pause during approach. Track regions are color-coded as in Fig. 1. Dashed lines show rat trajectory with curved turnaround representing the rapid change in direction during a retreat and the horizontal line representing a stationary pause during an MTA. Feeders are shown at either end and grey blocks represent the doorway. White circles highlight epoch of interest. b. Top: Decoded position aligned to turnaround during Retreats (left), MTAs (center), and matched inbound control laps (right). Window is shown for 2 seconds on either side of the event. Black vertical line represents the event of interest, in this case the moment inbound travel begins. Horizontal lines show landmark locations (Doorway; light grey, Attack threshold; orange, robot; dark grey). Solid lines represent the landmark position for cohort 1, and dashed for cohort 2. Bottom: Corresponding time-frequency power spectrograms. c. Power spectral density (PSD) curves (1-300 Hz) for Retreat Inbound (IB; orange), MTAs IB (brown), and Control IB (grey), averaged across rats (cohorts combined). Inset shows 1-25Hz range of the same data. d. Difference in decoded position during inbound return (Retreat IB - MTA IB). Positions were spatially warped to account for differing journey lengths. Cohort 1 is shown on the left and cohort 2 on the right. Green indicates stronger representation in the first condition of the subtraction (Retreat IB), whereas purple indicates stronger representation in the second condition (MTA IB). Horizontal lines denote key landmarks. e. Mean decoding bias (Retreat IB - MTA IB) across track regions (cohorts combined). Positive values indicate stronger representation during the first condition in the subtraction, whereas negative values indicate stronger representation during the second condition.

Comparison of pauses preceding abort and continue.

a. Schematic of pause preceding an abort (MTA; left) and continuing on to the far feeder (MTC; right). Track regions are color-coded as in Fig. 1. Dashed lines show rat trajectory. Feeders are shown at either end and grey blocks represent the doorway. White circles highlight epoch of interest. b. Top: Decoded position aligned to middle of pause period in MTAs (left), MTCs (right). Window is shown for 2 seconds on either side of the event. Black vertical line represents the event of interest. Horizontal lines show landmark locations for cohort 1 (solid), and cohort 2 (dashed). Bottom: Corresponding time-frequency power spectrograms. c. Power spectral density (PSD) curves (1-300 Hz) for MTA pause (brown), MTC pause (pink), averaged across rats (cohorts combined). Inset shows 1-25Hz range of the same data. d. Difference in decoded position during pause period (MTA Pause - MTC Pause). Positions were temporally warped to account for differing pause durations. Cohort 1 is shown on the left and cohort 2 on the right. Green indicates stronger representation during the MTA pause, whereas purple indicates stronger representation during the MTC pause. Horizontal lines denote key landmarks. e. Mean decoding bias (MTA Pause - MTC Pause) across track regions (cohorts combined). Positive values indicate stronger representation during the MTA Pause, whereas negative values indicate stronger representation during the MTC Pause.

Comparison of outbound journeys preceding pause.

a. Schematic of outbound journey preceding the pause associated with an MTA (left) and MTC (right). Track regions are color-coded as in Fig. 1. Dashed lines show rat trajectory. Feeders are shown at either end and grey blocks represent the doorway. White circles highlight epoch of interest. b. Top: Decoded position aligned to beginning of outbound journey associated with MTAs (left), MTCs (middle), and control laps (right). Window is shown for 1 second before doorway crossing to 3 seconds after. Black vertical line represents the event of interest. Horizontal lines show landmark locations for cohort 1 (solid), and cohort 2 (dashed). Bottom: Corresponding time-frequency power spectrograms. c. Power spectral density (PSD) curves (1-300 Hz) for MTA outbound (OB; brown), MTC OB (pink), and control OB laps (grey) averaged across rats (cohorts combined). Inset shows 1-25Hz range of the same data. d. Difference in decoded position during outbound journey (MTA OB - MTC OB). Positions were spatially warped to account for differing journey lengths. Cohort 1 is shown on the left and cohort 2 on the right. Green indicates stronger representation during the MTA OB, whereas purple indicates stronger representation during the MTC OB. Horizontal lines denote key landmarks. e. Mean decoding bias (MTA OB - MTC OB) across track regions (cohorts combined). Positive values indicate stronger representation during the MTA OB, whereas negative values indicate stronger representation during the MTC OB.