STRATA
Skip to study
XLVII

SOURCE RECORD / 1982

Memory as a collective state

DocumentedR01 · 2026-11-18
Scope of this record

KNOWN refers to the model described in the abstract, not the claim that all biological memory or cognition follows this architecture.

Statement

Hopfield describes a network model in which collective dynamics support content-addressable memory: a partial cue can lead toward a stored pattern. [source]

Equation

Unrecorded

The connection is scoped to this source. No broader cross-domain equality is established by this entry.

Observation

The abstract connects simple processing elements with emergent computational behavior. Its target is a model of associative recall rather than a complete account of intelligence. [source]

Progressive questions

Which patterns are stored and under what update rule? How much cue corruption can be tolerated? Which states attract the dynamics but were never intended as memories? What changes when the input distribution shifts?

Source response

The paper presents asynchronous collective processing as a way to implement content-addressable memory. Performance limits and exact energy conditions require the full specification before a faithful reproduction. [source]

Interpretation

XLVII proposal: visualize state transitions and recalled patterns, not a glowing brain presented as proof. An apparent convergence is meaningful only relative to the stored set and the stated retrieval criterion.

Conclusion

The surviving mechanism to investigate is collective associative recall within a defined model. [source]

Next experiment

Proposed: implement a small explicitly specified network, record stored patterns and corrupted cues, and measure retrieval error across corruption levels. Include unintended attractors and update-order tests. Do not call a newly written implementation a reproduction until its assumptions match the original.

Work through a related example.

III.03 / Information-processing cognition

Fit two explanations to the same trend.

Construct illustrative response times for memory sets of two, four, and six items. Fit a straight line and identify its slope and intercept. Then ask whether increasing perceptual difficulty could change the intercept while a scanning process changes the slope; a plot becomes useful when it separates those predictions.

III.04 / Neuroscience of cognition

Compare repeated responses to the same stimulus.

Select one stimulus repeated across trials and plot the response of one recorded unit or imaging trace each time. Overlay the average only after displaying the variation. Then compare a second stimulus; a difference between averages becomes more interpretable when its size is seen against the within-condition spread.

IV.02 / McCulloch–Pitts models and early artificial neurons

Build an AND gate from a threshold unit.

Let two binary inputs each have weight one and let the unit fire when their sum reaches two. Write all four input combinations and the resulting output. Lower the threshold to one and the same weights implement OR. This isolates the role of the threshold before adding learning or continuous-valued activations.

Follow earlier and later work.

Before

  • Proposed connection · UNKNOWNInformation is not meaning

    Editorial research connection to investigate; not a verified correspondence, causal lineage, or mathematical equivalence.

After

  • Proposed connection · UNKNOWNPrediction and residual error

    Editorial research connection to investigate; not a verified correspondence, causal lineage, or mathematical equivalence.

Follow the chain
  1. Information is not meaning · 1 link away · UNKNOWN
  2. Prediction and residual error · 1 link away · UNKNOWN
  3. Compression → classification · 2 links away · UNKNOWN
  4. Erasure has conditions · 2 links away · UNKNOWN
  5. Attention without recurrence · 2 links away · UNKNOWN
  6. The chemical compass · 2 links away · UNKNOWN
  7. Dynamic Markov modelling · 3 links away · KNOWN
  8. Compression, reopened · 3 links away · UNKNOWN
  9. Reversibility and the record · 3 links away · UNKNOWN
  10. Compute has a budget · 3 links away · UNKNOWN
  11. What survives a replication? · 4 links away · UNKNOWN

Placement.

Technology

This entry records the cited source and a proposed next investigation. It makes no claim of historical abandonment, suppression, or independent reproduction.

Sources.

  1. Neural networks and physical systems with emergent collective computational abilities

    J. J. Hopfield / Proceedings of the National Academy of Sciences · 79(8), 2554–2558; DOI 10.1073/pnas.79.8.2554; abstract.

    Abstract and bibliographic metadata inspected; scanned full-paper equations not transcribed. Checked 2026-11-18.

Record history
  1. 2026-11-18 / R1 / SOURCE_SEEDED

    Source-limited editorial seed with explicit inspection scope. Research questions and proposed experiments are not executed results.