The rat stopped. The sequence ran backwards.

We often picture memory as a recording we can play again. This study makes that picture less straightforward.

What supports the answer?

  1. Reported finding: Reverse replay appeared during awake pauses
    Reported finding

    Place-cell activation sequences recurred in reverse after running, compressed into hundreds of milliseconds.

    Scope and conditions: Four rats, with familiar and new track sessions; recordings in the hippocampus.

    Source passage · 2006

    Read the recording setup and sequence comparison; the PDF uses its own page numbering, 1–4.

    PDF p. 1; Fig. 2 on p. 2; Figs. 3–4 and Methods on pp. 3–4. About the source and its version
  2. Method described: The comparison used place-cell sequences
    Method described

    The researchers compared activity during pauses with sequences defined by cells’ firing locations during running, and tested against shuffled cell orders.

    Scope and conditions: A neural recording experiment, not a report of subjective experience.

    Source passage · 2006

    Read the recording setup and sequence comparison; the PDF uses its own page numbering, 1–4.

    PDF p. 1; Fig. 2 on p. 2; Figs. 3–4 and Methods on pp. 3–4. About the source and its version

Where this explanation stops.

  • The experiment recorded neural activity in rats; it cannot tell us what they consciously remembered.
  • The proposed role in learning was not established by selectively disrupting replay in this experiment.
  • This is one historical finding, not a current review of neuroscience or a recipe for chatbot memory.

What to take with you

Atlas interpretation: ask what reusing an experience might accomplish, then ask how that proposed function could be tested.

Follow the answer back to its sources.

Peer-reviewed research · Source date: 12 February 2006

Reverse replay of behavioural sequences in hippocampal place cells during the awake state

David J. Foster, Matthew A. Wilson

Nature 440, 680–683 · Published online article, four-page PDF hosted in an MIT course archive

Online publication: 12 February 2006. Journal issue: 30 March 2006. The linked PDF is numbered 1–4; journal pagination is 680–683.

DOI: 10.1038/nature04587

Source passages, license and conditions ↗

How this version was checked

Human-reviewed by Daniel · 22 October 2026. The approval applies to revision 2, its text, visuals and source references.

Why this piece

This gives a concrete counterexample to the idea of memory as a faithful chronological playback. The path into AI asks what can be borrowed from a phenomenon without treating two systems as identical.

Human-reviewed by Daniel · Ask the editor

What changed, and why.

This history comes from saved content versions. A new revision can be an editorial change; it does not by itself mean that the underlying facts changed.

Latest reviewed revision

Correction · revision 2

Corrected the study year to refer to publication rather than recording, and clarified page references for Figures 3–4 and Methods. The research claims and limitations are unchanged.

Recorded 22 October 2026 · reviewed 22 October 2026 · Human-reviewed by Daniel

Compare revision 1 with 2

The wording below is taken from those exact saved versions.

Scope

Before · revision 1

Archival research · 2006 recordings in four rats. This is not a finding about human conscious recollection.

After · revision 2

Archival research · Study published in 2006, involving four rats. This is not a finding about human conscious recollection.

Source references: Reverse replay appeared during awake pauses

Before · revision 1

Reverse replay of behavioural sequences in hippocampal place cells during the awake state · Published online article, four-page PDF hosted in an MIT course archive · PDF pp. 1–2; Figures 2–4 and Methods · https://courses.csail.mit.edu/6.S079/pdf/2006%20Wilson.pdf#page=1

After · revision 2

Reverse replay of behavioural sequences in hippocampal place cells during the awake state · Published online article, four-page PDF hosted in an MIT course archive · PDF p. 1; Fig. 2 on p. 2; Figs. 3–4 and Methods on pp. 3–4. · https://courses.csail.mit.edu/6.S079/pdf/2006%20Wilson.pdf#page=1

Source references: The comparison used place-cell sequences

Before · revision 1

Reverse replay of behavioural sequences in hippocampal place cells during the awake state · Published online article, four-page PDF hosted in an MIT course archive · PDF pp. 1–2; Figures 2–4 and Methods · https://courses.csail.mit.edu/6.S079/pdf/2006%20Wilson.pdf#page=1

After · revision 2

Reverse replay of behavioural sequences in hippocampal place cells during the awake state · Published online article, four-page PDF hosted in an MIT course archive · PDF p. 1; Fig. 2 on p. 2; Figs. 3–4 and Methods on pp. 3–4. · https://courses.csail.mit.edu/6.S079/pdf/2006%20Wilson.pdf#page=1

Sources used before

Sources used after

Available version records (2)
  1. Revision 2 · Correction · reviewed

    Corrected the study year to refer to publication rather than recording, and clarified page references for Figures 3–4 and Methods. The research claims and limitations are unchanged.

    Recorded 22 October 2026 · reviewed 22 October 2026 · Human-reviewed by Daniel

    Answer in this version

    Atlas keeps a result and a possible explanation on different footing. A plausible story should tell us what evidence to look for next.

    What remained unresolved

    A stronger test would need to manipulate replay and measure learning, separating it from other activity. This discovery stops at the 2006 observation and proposal; it does not claim to summarize later evidence.

    Takeaway and scope

    The order of experience and the order of replay can differ.

    Archival research · Study published in 2006, involving four rats. This is not a finding about human conscious recollection.

    Reverse replay appeared during awake pauses

    Place-cell activation sequences recurred in reverse after running, compressed into hundreds of milliseconds.

    Four rats, with familiar and new track sessions; recordings in the hippocampus.

    The comparison used place-cell sequences

    The researchers compared activity during pauses with sequences defined by cells’ firing locations during running, and tested against shuffled cell orders.

    A neural recording experiment, not a report of subjective experience.

    Learning is a proposed function here

    The authors propose that reverse replay could help connect a later reward to preceding events.

    A proposed explanation, not a causal finding of this experiment.

    Sources in this version

  2. Revision 1 · First recorded version · reviewed

    First source-checked editorial draft on awake reverse replay. Observation and proposed function are kept separate. Awaiting human review.

    Recorded 22 October 2026 · reviewed 22 October 2026 · Human-reviewed by Daniel

    Answer in this version

    Atlas keeps a result and a possible explanation on different footing. A plausible story should tell us what evidence to look for next.

    What remained unresolved

    A stronger test would need to manipulate replay and measure learning, separating it from other activity. This discovery stops at the 2006 observation and proposal; it does not claim to summarize later evidence.

    Takeaway and scope

    The order of experience and the order of replay can differ.

    Archival research · 2006 recordings in four rats. This is not a finding about human conscious recollection.

    Reverse replay appeared during awake pauses

    Place-cell activation sequences recurred in reverse after running, compressed into hundreds of milliseconds.

    Four rats, with familiar and new track sessions; recordings in the hippocampus.

    The comparison used place-cell sequences

    The researchers compared activity during pauses with sequences defined by cells’ firing locations during running, and tested against shuffled cell orders.

    A neural recording experiment, not a report of subjective experience.

    Learning is a proposed function here

    The authors propose that reverse replay could help connect a later reward to preceding events.

    A proposed explanation, not a causal finding of this experiment.

    Sources in this version