Hippocampal replay is one of the cleanest demonstrations that the brain revisits its own experience. During sleep or quiet pauses, place cells fire in compressed sequences that recapitulate the paths an animal recently traveled. In laboratory mazes — typically a meter or two — replays sweep through most of the environment. The sequence starts near one end and traverses to the other, compressing minutes of real navigation into hundreds of milliseconds. The replay looks like a fast-forwarded movie of the journey.
Eliav, Maimon, Sarel, Palgi, Las, and Ulanovsky (Cell, 2025) recorded hippocampal replay in bats flying through a 200-meter tunnel — a hundred times longer than a standard maze. The replays were there. The compression was there. Individual place cells fired in sequence, respecting the spatial order of their fields. The mechanism works.
But the replays covered only about 6% of the tunnel.
Each replay event depicted a short fragment of the trajectory — roughly 12 meters of a 200-meter space. Not a sweep from end to end. Not a summary. A piece. During sleep, the bat's hippocampus replayed hundreds of these fragments, each one a different section, but no single replay spanned more than a small fraction of the total route.
In small environments, this constraint is invisible. If replay naturally covers about 12 meters and the maze is 2 meters long, the replay spans the whole thing. The finding looks like a complete traversal. Scale the environment up by a factor of a hundred, and what looked like a property of the replay — completeness — is revealed as an artifact of the arena's size. The replay was always fragmented. The fragments just happened to be longer than the maze.
The structural observation: the apparatus that looked like it captured the whole experience was never designed to do that. It captured what fit within its reach. At small scale, its reach exceeded the environment, producing the illusion of totality. At large scale, the boundary became visible.
The authors suggest the fragmentation may serve memory chunking — breaking large experiences into digestible pieces for consolidation. That's plausible. But the sharper finding is simpler: the system has a fixed spatial bandwidth. What changes with environment size is not the replay mechanism but the fraction of reality it can represent in any single pass.