EPISODIC MAP OF A COGNITIVE TRAJECTORY

Episodic memories are often assumed to be encoded by stable neuronal assemblies or persistent firing patterns, yet population-level recordings frequently reveal strong episode decoding in the absence of obvious single-neuron correlates, raising questions about where episodic identity resides. We addressed this problem using a task design that generates recurrent episodes matched in behavioral and cognitive state, yet dissociated in space and time, enabling isolation of episode-specific neural structure beyond trivial contextual similarity. Analyzing hippocampal population activity in freely behaving rodents, we found that target episodes were associated with highly consistent, directional population trajectories that evolved reproducibly across repetitions. Across episodes, population activity progressively converged toward constrained regions of state space (“funneling”), indicating that animals were repeatedly driven through similar population-level neural configurations despite differences in spatial location and temporal context, whereas randomly sampled control episodes showed absent trajectory organization. To dissociate task-related population dynamics from contextual influences, we residualized neural activity with respect to movement, spatial, and temporal variables. This procedure strongly reduced similarity between episodes by removing shared variance of these covariates, abolishing trivial sources of population resemblance; importantly, trajectory directionality and convergence remained robust, demonstrating that these features reflect intrinsic task-related dynamics rather than behavioral confounds. Despite the absence of persistent episode-specific neurons, episode identity remained decodable from brief population snapshots. Together, these results indicate that episodic identity is not carried by stable single units or assemblies, but instead emerges from distributed, high-dimensional population states evolving along reproducible trajectories on top of a slowly drifting hippocampal manifold.