Hippocampal representations differentiate reactive and anticipatory responses during foraging under threat

Adaptive behavior under threat requires balancing reward pursuit against the risk of harm. During approach-avoidance conflict, animals often pause at decision points, but whether these pauses reflect a unified process or distinct decision states remains unclear. Here, we replicate and extend findings from Calvin et al.¹ by analyzing hippocampal activity in rats performing a predator-based foraging task across two cohorts. We compared three behaviors: mid-track aborts (MTAs), mid-track continues (MTCs), and attack-triggered retreats. Behaviorally, MTAs and MTCs emerged from a shared pause state but led to different outcomes, whereas retreats reflected rapid, reactive escape following attack. Despite similar behavioral endpoints (return to safety), retreats and MTAs differed markedly in movement dynamics and neural activity. During retreats, hippocampal representations remained biased toward the attack location, consistent with ongoing encoding of immediate threat. In contrast, MTAs showed a shift in representation toward safe locations following the decision to abort. During pauses, hippocampal activity differentiated future behavioral outcomes before movement diverged: pauses preceding MTAs showed stronger representation of threat-related locations, whereas pauses preceding MTCs emphasized goal-related locations. These representational biases were already present during the outbound approach, indicating that decision-related processes emerged at the beginning of the outbound journey. Across experience, representations of threat and goal locations became increasingly differentiated when encountering an attacking robot, diminished during extinction, and re-emerged when the attack was introduced again. Together, these findings extend prior work by dissociating hippocampal representations associated with reactive escape from those underlying anticipatory, anxiety-like decision-making, suggesting that the hippocampus dynamically tracks behaviorally relevant features to guide decisions under threat.