Reconsolidation and the Reconstructive Nature of Memory: Implications for the Stability of Emotional Memories

Abstract

Consolidation theory long held that a memory trace, once stabilised by protein-synthesis-dependent processes, was resistant to disruption. The present review examines the evidence that retrieval returns consolidated memories to a labile state from which they must be restabilised (reconsolidation), and asks what this implies for the stability of emotional memories. We review the early demonstration by Misanin, Miller and Lewis (1968), the amygdala studies of Nader, Schafe and LeDoux (2000), the boundary conditions that determine whether reactivation destabilises a trace (reminder duration, trace age and strength, prediction error), and the human literature on pharmacological and behavioural interference with reconsolidation, including the contested replication record of the reactivation–extinction procedure. We then situate reconsolidation within the constructive-memory tradition of Bartlett, Loftus and Schacter. The main conclusion is that emotional memories are neither fixed records nor freely rewritable: they are updated under specific, partly characterised conditions, most reliably when retrieval carries a prediction error. For a programme aiming to preserve emotional memory, any elicitation of a memory is potentially an intervention on it, and preservation must be conceived as the documentation of dated reconstructions rather than the retrieval of an original.

Keywords: reconsolidation; constructive memory; fear conditioning; prediction error; propranolol; misinformation effect

1. Introduction

The standard model of memory formation distinguishes a brief post-encoding period during which a trace is vulnerable from a subsequent state in which it is stable. During the vulnerable period, amnesic agents such as electroconvulsive shock or protein-synthesis inhibitors abolish the memory; afterwards they are without effect. The implicit corollary of this consolidation hypothesis was that consolidation happens once: a trace was not thought to re-enter a state of vulnerability simply because it had been retrieved.

Two lines of research have challenged this corollary from opposite directions. The first, neurobiological, is the reconsolidation literature: animal and human work showing that retrieval can render a consolidated memory transiently labile, such that amnesic agents administered after reactivation impair subsequent expression. The second, psychological, is the constructive-memory tradition, which since Bartlett (1932) has documented that remembering is reconstructive, shaped by schemas, post-event information and current goals. These literatures converge on a single claim: the act of remembering is not neutral with respect to the memory remembered.

The present review summarises the reconsolidation evidence with attention to its boundary conditions, since the popular reading that every recall rewrites the past overstates what has been shown; assesses the human literature, where translational promise has been tempered by a mixed replication record; and draws out the implications for efforts to preserve emotional memories.

2. From consolidation to reconsolidation

2.1 Early evidence and its neglect

The first clear demonstration that a consolidated memory could be made vulnerable by reactivation is attributed to Misanin, Miller and Lewis (1968). Twenty-four hours after rats received a single tone–footshock pairing, beyond the consolidation window as then understood, one group was re-exposed to the tone and then given electroconvulsive shock; a second group received the shock without re-exposure. Only the reactivated group showed retrograde amnesia. The finding was replicated during the 1970s but was widely treated as a curiosity and largely set aside for three decades (Dudai, 2012).

2.2 Nader, Schafe and LeDoux (2000)

The modern literature dates from Nader, Schafe and LeDoux (2000). Rats were trained on auditory fear conditioning, in which freezing to a shock-paired tone indexes the memory. One day after training, the tone was presented once without shock, and the protein-synthesis inhibitor anisomycin was infused into the lateral and basal amygdala immediately afterwards. Animals tested the following day showed markedly reduced freezing relative to vehicle-infused controls, whereas anisomycin without reactivation, and reactivation without anisomycin, left the memory intact. The same result held for fourteen-day-old memories, ruling out a merely extended consolidation window. The authors concluded that retrieval returns the trace to a labile state and that restabilisation, or reconsolidation, requires new protein synthesis. The design logic underlies every subsequent study: reconsolidation is inferred from an interaction in which an amnesic manipulation impairs the memory only when combined with prior reactivation.

2.3 Dudai’s synthesis

Dudai’s reviews (2004, 2012) provided the conceptual framework in which these findings are discussed. He distinguished synaptic consolidation, over hours and dependent on local protein synthesis, from systems consolidation, over weeks to years and involving circuit reorganisation. Reconsolidation re-engages synaptic consolidation after reactivation, and its function is proposed to be adaptive: it permits an existing trace to be updated with information available at retrieval. Dudai (2012) was explicit that this reading remains a hypothesis: post-reactivation vulnerability is well established in several species, but the claim that the labile period exists to permit updating is an inference from it.

3. Boundary conditions

If every retrieval destabilised every memory, far more forgetting and distortion would be observed than is. The effect is gated by several conditions, and characterising them has been the principal work of the past two decades (Lee, Nader & Schiller, 2017).

3.1 Reminder duration

Pedreira and Maldonado (2003), working with the crab Chasmagnathus, showed that reminder duration determines which process is engaged. A brief re-exposure to the training context destabilised the memory, so that a protein-synthesis inhibitor produced amnesia; a prolonged re-exposure produced extinction, which the same inhibitor now blocked while leaving the original memory intact. Reconsolidation and extinction are thus alternative outcomes of reactivation, with the boundary set by the amount of non-reinforced exposure. The reminder must be long enough to destabilise the trace but not so long as to initiate extinction, and the optimal parameters are not known a priori.

3.2 Trace age and strength

In rodents, memories that are weeks old or overtrained often fail to show post-reactivation vulnerability under reminder conditions sufficient for younger or weaker memories, although more discrepant reminders can sometimes restore it (Lee et al., 2017). The generality of this constraint in humans is uncertain, since most human studies use memories that are one day old.

3.3 Prediction error

The most theoretically productive boundary condition is prediction error. Sevenster, Beckers and Kindt (2013) trained participants on differential fear conditioning and administered propranolol after a reactivation trial. When the reactivation trial was arranged so that its outcome was fully predicted, propranolol had no effect on later fear-potentiated startle; when reactivation involved a violation of expectancy, propranolol abolished the startle response. Retrieval alone is therefore not sufficient: current input must diverge from what the trace predicts. This is consistent with Dudai’s functional hypothesis, since a memory is opened for revision precisely when there is something to revise, and it constrains applications, because the same reminder can produce a prediction error for one individual and none for another.

Table 1. Principal boundary conditions on memory destabilisation.

4. Reconsolidation in humans

4.1 Pharmacological interference

Kindt, Soeter and Vervliet (2009) provided the first widely cited human demonstration. After differential fear conditioning, one group received oral propranolol, a β-adrenergic antagonist, before a single unreinforced presentation of the conditioned stimulus on the following day. On the third day, fear-potentiated startle was abolished in this group but not in placebo or no-reactivation controls, and did not return after reinstatement shocks. Declarative knowledge of the contingency, indexed by expectancy ratings, was unaffected. Whether this constitutes erasure of an amygdala-dependent component or a change in its expression remains debated (Lee et al., 2017), but the dissociation between affective expression and declarative content has been replicated within the same laboratory.

4.2 Behavioural updating and its replication record

Schiller et al. (2010) asked whether the reconsolidation window could be exploited without drugs. After fear conditioning, participants received a single reminder trial and then underwent extinction either ten minutes later, within the putative window, or six hours later, outside it. Return of skin-conductance responses on the third day was absent only in the ten-minute group, and a subset showed no return of fear approximately one year later. Extinction inside the window was interpreted as updating the fear memory itself, rather than forming a separate inhibitory trace.

The replication record of this reactivation–extinction effect is mixed, and the present review regards it as contested. Some laboratories reported conceptual replications; others failed to observe the effect with startle or skin conductance (Lee et al., 2017). Chalkia et al. (2020) conducted a registered, high-powered direct replication of the original design and found no persistent attenuation of fear in the ten-minute group, with return of fear across all conditions. They concluded not that reconsolidation does not occur in humans, but that the procedure as described does not reliably produce it. The most parsimonious reading is that behavioural updating is possible but sensitive to parameters not yet controlled, plausibly including whether the reminder generates a prediction error (Sevenster et al., 2013).

4.3 Episodic memory

Hupbach, Gomez, Hardt and Nadel (2007) extended the logic to episodic memory. Participants learned a list of objects on day one. On day two, one group received a subtle reminder (the same experimenter, the same room and a question about the original session) before learning a second list; another group learned it without the reminder. On day three, recall of the first list showed substantial, asymmetric intrusions from the second list only in the reminded group, as a reconsolidation account predicts. The phenomenon is thus not confined to amygdala-dependent aversive learning.

5. Constructive memory: the psychological tradition

5.1 Bartlett

Bartlett (1932) asked participants to reproduce an unfamiliar folk narrative, “The War of the Ghosts”, at intervals over weeks and months. Successive reproductions became shorter, more coherent by the participants’ cultural standards, and populated with details absent from the original. He concluded that remembering is a reconstruction from fragments guided by schemas rather than the re-reading of a stored record. His methods were informal by contemporary standards, but the central claim has survived a century of testing.

5.2 Loftus and the misinformation effect

Loftus (2005) reviewed three decades of research on the misinformation effect, in which post-event information alters subsequent recall. A reliable proportion of participants exposed to misleading post-event information incorporate it into their reports. Loftus also reviewed studies in which fictitious childhood events, such as being lost in a shopping mall, were implanted through suggestion in roughly a quarter of participants. The paradigm structurally resembles the reconsolidation designs reviewed above: a memory is reactivated, new information follows, and that information is later reported as part of the original. It would be premature to assert that misinformation effects are mediated by synaptic reconsolidation, but the convergence with Hupbach et al. (2007) is at minimum suggestive.

5.3 Schacter’s adaptive account

Schacter (1999) organised memory failures into seven categories: sins of omission (transience, absent-mindedness, blocking), sins of commission (misattribution, suggestibility, bias) and persistence, the unwanted recurrence of distressing memories. His argument, developed in Schacter (2012), is that these failures are by-products of processes that are adaptive on the whole: a system that extracts gist, integrates new information and simulates future events will necessarily be prone to distortion. On this view, reconsolidation is a candidate mechanism through which such constructive processes are implemented, although the neurobiological and functional accounts are argued at different levels and neither is fully specified.

6. Implications for emotion preservation

The literature reviewed yields several constraints on any programme that aims to preserve human emotional memory. We formulate them as testable propositions.

First, there is no stable original to be retrieved. An emotional memory retrieved and reconsolidated many times is the product of that history, and its current form cannot be distinguished, from the inside, from earlier forms. Preservation should therefore be conceived as the documentation of dated reconstructions. A record of a grief one month after loss and a record of the same grief ten years later are different objects, and a programme that treats them as interchangeable misdescribes both.

Second, elicitation is intervention. Any procedure that asks a person to retrieve an emotional memory in order to document it reactivates the trace and, under some conditions, destabilises it. Those conditions include a brief reminder and, most reliably, a prediction error. Elicitation contexts that are novel, ask unexpected questions, or confront the person with discrepant information are the ones most likely to modify what they aim to record. This is testable: memories elicited under high-discrepancy conditions should show greater change on subsequent elicitation than memories elicited under low-discrepancy conditions.

Third, emotional and declarative components may dissociate. The propranolol studies show that affective expression can be attenuated while declarative content is spared. A programme that documents only declarative content risks preserving a description of an emotion rather than the emotion; one that documents only physiological markers risks a response detached from its object.

Fourth, the replication record counsels modesty. Human reconsolidation research has established that updating is possible but not that it is reliable, and the boundary conditions are less well characterised in humans than in rodents. Preservation practices should measure change rather than presume its absence; current evidence requires caution but does not specify safe procedures.

7. Conclusion

Reconsolidation research has replaced a model in which consolidation happens once with one in which retrieval can, under specific conditions, reopen a trace for revision. The evidence for post-reactivation vulnerability is strong in animals and reasonably strong for pharmacological interference in humans; the evidence for behavioural updating in humans is contested, and prediction error emerges as the most consistent gating factor. Read alongside the constructive-memory tradition, these findings support a view of emotional memory as a maintained reconstruction rather than a stored record. That view does not make emotional memories arbitrary; it makes them historical, and places the act of remembering, including remembering for preservation, among the forces that shape what is remembered.

References

Bartlett, F. C. (1932). Remembering: A study in experimental and social psychology. Cambridge University Press.

Chalkia, A., Schroyens, N., Leng, L., Vanhasbroeck, N., Zenses, A.-K., Van Oudenhove, L., Luyten, L., & Beckers, T. (2020). No persistent attenuation of fear memories in humans: A registered replication of the reactivation-extinction effect. Cortex, 129, 496–509.

Dudai, Y. (2004). The neurobiology of consolidations, or, how stable is the engram? Annual Review of Psychology, 55, 51–86.

Dudai, Y. (2012). The restless engram: Consolidations never end. Annual Review of Neuroscience, 35, 227–247.

Hupbach, A., Gomez, R., Hardt, O., & Nadel, L. (2007). Reconsolidation of episodic memories: A subtle reminder triggers integration of new information. Learning & Memory, 14(1–2), 47–53.

Kindt, M., Soeter, M., & Vervliet, B. (2009). Beyond extinction: Erasing human fear responses and preventing the return of fear. Nature Neuroscience, 12(3), 256–258.

Lee, J. L. C., Nader, K., & Schiller, D. (2017). An update on memory reconsolidation updating. Trends in Cognitive Sciences, 21(7), 531–545.

Loftus, E. F. (2005). Planting misinformation in the human mind: A 30-year investigation of the malleability of memory. Learning & Memory, 12(4), 361–366.

Misanin, J. R., Miller, R. R., & Lewis, D. J. (1968). Retrograde amnesia produced by electroconvulsive shock after reactivation of a consolidated memory trace. Science, 160(3827), 554–555.

Nader, K., Schafe, G. E., & LeDoux, J. E. (2000). Fear memories require protein synthesis in the amygdala for reconsolidation after retrieval. Nature, 406(6797), 722–726.

Pedreira, M. E., & Maldonado, H. (2003). Protein synthesis subserves reconsolidation or extinction depending on reminder duration. Neuron, 38(6), 863–869.

Schacter, D. L. (1999). The seven sins of memory: Insights from psychology and cognitive neuroscience. American Psychologist, 54(3), 182–203.

Schacter, D. L. (2012). Adaptive constructive processes and the future of memory. American Psychologist, 67(8), 603–613.

Schiller, D., Monfils, M.-H., Raio, C. M., Johnson, D. C., LeDoux, J. E., & Phelps, E. A. (2010). Preventing the return of fear in humans using reconsolidation update mechanisms. Nature, 463(7277), 49–53.

Sevenster, D., Beckers, T., & Kindt, M. (2013). Prediction error governs pharmacologically induced amnesia for learned fear. Science, 339(6121), 830–833.