A theory-guided model maps four recurrent modes of fear-goal arbitration across learning and relapse in rats
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
When defensive behavior declines, has the individual recovered? Addressing this translational question requires testing whether exploration and goal pursuit also return, and whether those gains survive time and context. We adapted rat fear-conditioning, extinction, delay, and contextual relapse paradigms to this problem. Male rats first explored the chambers under safer conditions and learned to press for food; aversive conditioning then reorganized the repertoire around danger. Extending tri-reference-point (TRP) theory from risky choice, we framed the repertoire around danger, the current condition, and goals. Across 47,004 five-second windows from 29 animals and 213 subject-sessions, we aligned freezing, 22- and 50-kHz calls, and lever presses. Subject-grouped, discrete-aware validation selected a diagonal Gaussian hidden Markov model at theory-guided K=4, decoding recurrent Fear/Danger, Quiescent, Exploratory, and Goal-directed modes. K=8 predicted held-out data better and subdivided these functional regions. Conditioning shifted occupancy from Exploratory toward Fear/Danger. By the final decile of initial extinction, Fear/Danger had fallen to 2.7%, yet Quiescent occupied 95.8%, with Exploratory at 0.6% and Goal-directed at 1.0%: fear reduction without broader functional restoration. Across related cohorts, later tests revealed exploratory-rich and renewed-defense trajectories. Goal-to-Fear transitions were depleted under two null models, and Fear/Danger occupancy was associated with fecal boli after session adjustment (bias-reduced p=0.036). Count-tailored models preserved temporal dependence but regrouped behavior, showing that state identity depends partly on representation. Together, multimodal measurement, TRP-derived landmarks, and temporal modeling make functional recovery testable as the history-dependent restoration of an individual’s behavioral repertoire, not merely the suppression of defense.
Plain-language summary
A rat may stop freezing when a once-dangerous cue is repeatedly presented without harm, yet remain withdrawn and fail to explore or pursue a learned reward. Functional recovery therefore cannot be judged from reduced defense alone; it also requires asking whether the animal re-engages with its environment, resumes effective goal pursuit, and maintains those gains across time and context. We designed the rat studies around this translational distinction. Before danger was introduced, the rats encountered the chambers under safer conditions, explored, and learned that pressing a lever earned food. Footshock then made the situation dangerous, shifting behavior from engagement and goal pursuit toward defense. Extinction, delay, and changes of context allowed us to ask not simply whether freezing declined, but what happened to the wider repertoire and whether renewed function endured or defense returned. We measured freezing, 22-kHz alarm calls, 50-kHz calls that occur in social, rewarding, and novel situations, and food-reinforced lever presses together in five-second windows. A hidden Markov model then decoded a moment-to-moment path for each rat-session through four recurring patterns: Fear/Danger, Quiescent, Exploratory, and Goal-directed. Conditioning shifted the repertoire from Exploratory toward Fear/Danger. During the first extinction session, Fear/Danger later declined, yet Quiescent dominated rather than Exploratory or Goal-directed, and a change of context could restore defense. This is the key distinction: reduced defense, low-output quietness, renewed environmental engagement, and recovered task performance are different outcomes. The pooled percentages for the four modes summarize an unequal experimental schedule and are not normal or desirable targets. Models with more states refined the broad regions, while count-tailored models grouped some moments differently, so the exact classification stays model-dependent. Fear/Danger occupancy was also associated with fecal boli, an autonomic measure kept out of fitting. The framework therefore traces functional recovery as a history-dependent return of the wider repertoire, not simply a fall in defensive behavior; prospective studies must still test whether individual trajectories predict durable recovery or relapse.
Impact statement
A theory-guided temporal model traces how aversive learning, extinction, delay, and context redistribute defensive, quiescent, exploratory, and goal-directed functioning, distinguishing fear reduction from broader recovery.
