Human memory reconsolidation can be explained using the temporal context model

Per B. Sederberg, Samuel J. Gershman, Sean M. Polyn, Kenneth A. NormanPsychonomic 2011 · DOI

This 2011 paper from Psychonomic Bulletin & Review provides a compelling "Occam’s Razor" argument for one of the most debated phenomena in memory research: reconsolidation. While many researchers argue that retrieving a memory physically "unlocks" it at a molecular level to allow for updates, Sederberg and colleagues demonstrate that a well-established cognitive model can explain these effects through simple contextual associations—no protein synthesis inhibitors required.

The Reconsolidation Challenge

In the early 2000s, animal studies suggested that when a memory is retrieved, it enters a "labile" (malleable) state where it can be changed, strengthened, or even erased. This is known as reconsolidation. In humans, researchers like Hupbach et al. (2007) demonstrated this behaviorally: if you learn List A on Monday, get a "reminder" of List A on Wednesday, and then immediately learn List B, you will later "intrude" many List B items when trying to recall List A. Curiously, this effect is asymmetric—you don't intrude List A items when recalling List B.

Hupbach argued this proved the reminder "unlocked" List A, allowing List B items to be integrated into it. Sederberg et al. challenge this "cellular update" view, asking if existing models of how we navigate time and context can explain the same data more simply.

Methods: Modeling Contextual Drift

The authors employed the Temporal Context Model (TCM), specifically a version called TCM-A which includes a "leaky-accumulator" decision process. In TCM, "context" is a mathematical representation of your mental state that drifts slowly over time. When you learn an item, it is associated with the current context; when you retrieve an item, it reinstates the context in which it was learned.

To test their hypothesis, the authors mapped the experimental protocol of Hupbach et al. (2007) directly into simulation steps, as detailed in Figure 1.

Mapping of the Hupbach et al. (2007) experimental protocol to the specific computational steps in the TCM simulation.

The Core Mechanics:

  • Contextual Reinstatement: The "reminder" doesn't make the old memory malleable; it simply brings the old "Monday context" into the "Wednesday session." Because the Monday context is active while learning List B, List B items become strongly associated with it.
  • Recall-to-Reject: On immediate tests, the model compares the retrieved context of an item to the "current" context. If they match (i.e., "I just learned this today"), the model can reject the item if it's looking for an older list. This explains why intrusions are absent when testing occurs immediately after learning List B.

Key Results

The model achieved a very close fit to the behavioral data, with a Root Mean Square Deviation (RMSD) of only 1.71% for the 2007 recall data. As shown in Figure 2, the model perfectly captured the specific pattern where List B items are misattributed to List A in the reminder condition, but not the other way around (the asymmetry).

Comparison of behavioral data from Hupbach et al. (2007) and the TCM-A model simulations across different recall conditions.

Why the Asymmetry?

List B items are linked to the List A context because of the reminder, but List A items were never linked to the List B context (which didn't exist yet on Monday). Therefore, cuing with List A context pulls up both lists, but cuing with List B context only pulls up List B. By Day 3, the "today" signal that allowed participants to reject intrusions on Day 2 has faded, leading to the observed errors.

Conclusion

The authors demonstrated that "reconsolidation" effects can emerge from standard associative learning principles without needing to invoke specialized molecular "un-learning" or "updating" mechanisms. While the model has limitations regarding spatial specificity and strategic control, it offers a parsimonious explanation for how mental context navigates the passage of time and the interference of new information.

Summary updated

BibTeX
@inproceedings{sederberg2011human,
  author = {Per B. Sederberg and Samuel J. Gershman and Sean M. Polyn and Kenneth A. Norman},
  title = {{Human memory reconsolidation can be explained using the temporal context model}},
  booktitle = {Psychonomic},
  year = {2011},
  doi = {10.3758/s13423-011-0086-9},
}
Figure